Housing, consumption and borrowing: an assessment of recent personal sector behavior in the UK.

Pain, Nigel ; Westaway, Peter

1. Introduction

This note describes some results from recent research on the personal sector an the housing market in the UK, carried out at the National Institute over the last two years. The research has included the development of a theoretical micro-to-macro model which has subsequently been used to inform the specification of our econometric modelling work. This has resulted in the incorporation of a new system of equations for the personal sector in the NIESR macroeconometric model of the UK economy. This now includes new relationships characterising consumer spending (total and durables), housing demand, house prices, housing investment, consumer credit and mortgage equity withdrawal. These equations are closely related to those described in Pain and Westaway (1994) which were developed in a project commissioned by HM Treasury.

This note provides an overview of the main characteristics of the individual equations and how they relate to each other. Section 3 discusses some important theoretical issues and Section 4 summarises the empirical specification of the equations for consumers' expenditure, house prices and housing investment. Full details are given in NIESR (1994). Sections 5 and 6 concentrate on analysing th performance of the system as a whole, asking how well the equations fare at explaining the salient features of household behaviour during the 1980s and early-1990s. In particular, we focus on the role of financial liberalisation in changing the behaviour of consumer borrowing and attempt to quantify the extent to which these changes can explain the strength of the boom in the mid to late-1980s and the severity of the recession in the early-1990s.

2. Background

The behaviour of the personal sector has changed markedly over the past 15 years, coinciding with a period in which overall consumers' expenditure proved extremely difficult to forecast with any degree of accuracy (Britton and Pain, 1992). Growth in expenditure was consistently underestimated in the boom years of 1985 to 1988 and consistently over-estimated between 1989 and 1992 as the economy fell into recession.(1) The possible influence that these forecast errors had on policymaking decisions and hence on macroeconomic outturns, together with the associated concern that conventional models of personal secto behaviour had 'broken down', has prompted a re-examination of the determinants of personal sector expenditure and borrowing and the interrelationships between these decisions and developments in the housing market.

Borrowing by households rose rapidly in the 1980s, with the ratio of debt to personal disposable income doubling from 57 per cent at the end of 1980 to 116 per cent by 1990. Despite this, the overall net wealth of the personal sector, defined to include the value of housing and financial assets, rose to around 5 times income by the end of the decade from 3 1/4 times income at the start. Thi rise was helped by a rise in owner-occupancy and the buoyant behaviour of many asset prices. Annual house price inflation peaked at 28 per cent in 1988. The economy grew rapidly from the middle of the decade, with consumers' expenditure rising by 5 1/4 per cent per annum between 1984 and 1989. Investment expenditur on housing also rose sharply and the personal sector ran a financial deficit between 1987 and 1989, the first such period since the late-1950s.

These changes took place amidst widespread financial deregulation within the UK economy. Sargent (1991) and Kilpatrick (1991) both provide a detailed timetable of the main events in the liberalisation process. Two developments of particula importance were the abolition of quantitative controls on consumer credit in 1982 and the gradual relaxation of restrictions on mortgage lending. Deregulation at a time of rising house prices, allowed many households improved access to credit markets, with additional borrowing secured on previously unutilised housing collateral. The average loan offered by building societies t first-time buyers rose from some 76 per cent of the purchase price in 1980 to 8 per cent by 1986, see Chart 1. As Chart 2 illustrates, the level of both new (net) consumer credit and mortgage equity withdrawal--broadly the difference between new mortgage lending and investment in housing assets--rose sharply in the second half of the 1980s.

Monetary policy was tightened in the late-1980s as inflationary pressures began to emerge, with short-term interest rates rising from 8 per cent in early-1988 to 15 per cent by the end of 1989. This was followed by a marked change in the behaviour of the personal sector, with consumers' expenditure declining by over 2 per cent in 1991. The housing market weakened considerably and nominal prices fell during 1991-93, the first such decline for nearly forty years. New borrowing fell sharply, with total new borrowing in 1993 being equivalent to 3 3/4 per cent of disposable income, compared to 17 1/4 per cent at the height of the economic boom in 1988. In part this appears to be due to some degree of financial 're-regulation', with lenders adopting a more prudential attitude.

Chart 1 shows that the average loan-value ratio for first-time borrowers has recently declined to a level close to that prevailing immediately prior to the initial liberalisation of the mortgage market.

3. Theoretical issues

Our empirical work draws on the properties of a theoretical micro-to-macro mode of personal sector behaviour, originally developed by Ermisch and Westaway (1990,1994) and extended by Westaway (1993). In this model aggregate behaviour is derived by explicitly summing across overlapping generations of individual consumers who are all at different points in their respective life cycles. Individual behaviour itself is based on the conventional assumption that consumers allocate their lifetime income stream in order to maximise their utility. Two particularly important elements of the life-cycle model we use are the explicit focus put on the choice between consumer spending and expenditure on housing servicew and the assumption that bequests are passed between generations.

The primary use of microeconomic simulation models of this kind is to generate interesting macroeconomic insights by illustrating how different 'shocks' might be expected to impinge upon personal sector behaviour. A particular focus of ou work has been to illustrate the impact of financial liberalisation on individua behaviour and the effect on the dynamics of the response of aggregate consumption of a relaxation in borrowing gonstraints.

By way of illustration Chart 3 shows the response of aggregate consumption, relative to income, following a simulated change in borrowing constraints. In this example the permitted maximum level of debt for a particular individual ha been raised from 3 to 5 times labour income. (Precise details of all the parameter settings in the micro-to-macro model assumed for these simulations ar given in Westaway (1993).) An initial upwards jump in the consumption to income ratio of over 5 percentage points is gradually reversed, with the eventual outcome having a lower steady-state ratio than before. The temporary fall in th savings ratio occurs as extra borrowing opportunities are exploited. Eventually a new 'desired' debt-income ratio is attained. However at this point, for given labour income, interest payments are proportionately ligher than before and so aggregate consumption is reduced.

An interesting feature of this simulation is the protracted dynamic response, with the savings ratio (and the ratio of net wealth to income) only reaching a new steady state level over 50 years after the initial shock. This is caused purely by aggregation across age cohorts. Intuitively, at the time of the shock there are many individuals who have reached a stage in their life cycle at whic they are unaffected by the sudden removal of borrowing constraints.

The properties of theoretical models can also be used as benchmarks for the behaviour and specification of empirical models. These will often be less sophisticated because of inevitable data limitations due, for example, to difficulties in observing measures such as discount factors and the incidence o borrowing constraints.

Again by way of illustration it is of interest to consider the relationship between the demand for consumer goods and services and the consumption of housing services. Other things being equal, individuals can be expected to adjust their expenditures so that the marginal utility derived from consuming housing services is equal to that associated with consuming goods. In the presence of borrowing constraints this relationship will be distorted, with the nature and size of the distortion depending upon the form of the constraint. An important result obtained from the micro-to-macro model is that in the presence of a general constraint on borrowing (by an individual) the intratemporal condition linking demand for consumer goods and housing services is much less distorted by the constraint than the individual demands themselves. This provides a strong argument for modelling housing demand conditional on consumption itself. Doing so provides an explicit link between the permanent income measure implicit in consumers' expenditure and housing demand. Such a link could be lost if housing demand was simply conditioned on disposable incom and wealth.

Whilst this approach ensures a consistent theoretical treatment of consumption and housing demand, it remains necessary to account for the impact of financial deregulation on consumption. Three general approaches can be distinguished in the literature. The first, adopted in Muellbauer and Murphy (1989) and Darby an Ireland (1993), is to make some simplifying assumptions regarding the determinants of the consumption of credit-constrained households. The resulting expression, possibly allowing for time-varying parameters, can then be used in model of aggregate consumption. A second approach is to include dummy variables to proxy formal quantitative controls. For example, Melliss et al (1989) includ mortgage rationing and hire purchase dummies in their model of expenditure on consumer durables. Finally, it is possible to directly incorporate variables that embody the effects of financial conditions. An early example of this approach is provided by Cuthbertson (1980) who includes bank advances in a consumer durables equation. More recently, Carruth and Henley (1990, 1992), Miles (1992), Holmes (1993) and Patterson (1993) examine the impact of mortgage equity withdrawal on consumption and Westaway (1989) argues for the inclusion o consumer credit in the consumption equation.

All three approaches have some limitations, primarily because the incidence of borrowing constraints and the expenditure of credit-constrained households is unobserved. The first approach typically assumes that a credit-constrained household is one which is denied access to credit markets at all, rather than one which simply cannot achieve its desired level of debt. The second approach implies that borrowing constraints disappeared completely after deregulation. However, as the loan-value ratio shows, lenders have continued to exercise some degree of prudential restraint over the last few years. This implies that the incidence of credit constraints may vary over time, even in a deregulated, competitive financial market.

If credit constraints are being operated, whether for prudential reasons or because of regulatory controls, the supply of new credit might be considered an appropriate variable to include in the consumption function, with fluctuations in net advances of credit relative to disposable income providing an (inverse) indication of the fluctuations in credit constraints over time. In this light mortgage equity withdrawal can also be viewed as a means of attenuating a credi constraint (Patterson, 1993). (It should be emphasised that we are not suggesting that total borrowing be used, but merely that (small) proportion of it taken out for purposes other than housing investment and trade credit.)

Of course, if credit is purely demand determined, then observed changes in credit simply reflect the decision to consume. If so, the flow of credit and th resulting addition to consumption will be determined simultaneously, and 'explaining' consumption by means of movement in credit will be misleading. There are two means of testing this argument. First, formal statistical tests can be used to determine whether new credit is pre-determined (weakly exogenous with respect to consumers' expenditure. Results in Pain and Westaway (1994) suggest that this cannot be rejected. Second, and more informally, if consumer credit (and indeed mortgage equity withdrawal) is determined by the level of consumers' expenditure, then it should be possible to explain it using the conventional determinants of the consumption function. In practice, the work in Westaway (1989) and Pain and Westaway (1994) suggests that it is difficult to capture the determinants of credit using conventional variables, implying that credit is not purely demand determined.

These findings carry important implications for the modelling of personal secto behaviour, since they imply that direct use of the consumer credit and mortgage equity withdrawal series can be expected to yield more information of relevance to consumption than would be obtained by indirectly substituting their chosen determinants into the consumption function. We therefore adopt a 'disposable funds' approach to consumption, with disposable funds given by disposable incom plus the net flow of consumer credit and an estimated proportion of equity withdrawal.

4. Overview of empirical model

This section provides a broad overview of the equations actually estimated and how they relate to one another. There are three main parts to the overall model comprising the equations for consumer spending, credit and the housing market.

(i) Consumer spending

The long-run solution to the disposable funds model of consumption is shown as equation (1):

ln(C) = 0.951 ln(F + Y + 0.75 M) + 0.049 ln(W) - 0.0101 R (1)

Here C, Y, W and R denote total consumers' expenditure, disposable income, tota net wealth and interest rates respectively. F and M denote the constant price levels of consumer credit and equity withdrawal. Some three-quarters of equity withdrawal is estimated to eventually leak into consumption. Additional dynamic terms in unemployment are also included in order to capture precautionary savings effects. The long-run interest-rate effect is quite powerful, with a rise of 1 percentage point in interest rates eventually reducing consumption by 1 per cent, other things being equal.

Our equation also incorporates anticipatory effects through the inclusion of a lead term in expected consumption. Blake and Westaway (1993) show how such an equation can be derived from a forward-looking dynamic optimisation in the presence of adjustment costs. The equation implies that expectations of future income levels or future tax changes will have an explicit impact on the present level of expenditure. An equation of this form was used in our earlier analysis of the economic impact of the future tax changes announced in the two Budgets o 1993, see Pain (1994).

The split of total consumption between durables and non-durables expenditure is determined by an equation which explains the share of durables in aggregate expenditure. The main determinants used in Pain and Westaway (1994) are the proportion of consumption which is financed from credit, real interest rates an the relative price of durable goods.

(ii) The determination of credit

If credit constraints bite, then, as previously discussed, the equations for credit should include factors that affect the aggregate supply decision. The specification in place on the Institute model relates consumer credit to personal sector incomes, interest rates and unemployment. All three are likely to affect both the perceived credit risk of a particular credit applicant and the willingness of agents to expand their holdings of debt. Muellbauer and Murphy (1989) argue that credit constraints are likely to vary inversely with the nominal interest rate, since the probability of default rises as nominal debt service costs increase relative to nominal income.

Mortgage equity withdrawal (MEWD) arises from a number of different sources, some of which are structural, caused by demographic factors, and some of which are akin to consumer credit, see Holmans (1991) and Westaway (1994). The model we use attempts to capture these disparate influences in a single equation.(2) The main influences are found to be the stock of net housing equity, turnover i the housing market and the average loan-value ratio of first-time buyers. The non-linear form of our model implies that the impact of a particular change in lending criteria will be dependent on the existing stock of equity at the time of the change.

(iii) The housing sector

The proposed model of the housing market in Pain and Westaway (1994) has a number of similarities with previous studies in that equations are estimated fo the stock demand and flow supply of housing assets. House prices are assumed to behave in a market clearing fashion, moving so as to ultimately equalise the demand and supply of the stock of housing. However our equations also have a number of defining characteristics.

First, following the findings from the micro-to-macro model, the equation for housing demand is conditioned on the level of consumer spending. The ratio between the two is mainly determined by the user cost of housing, although in practice it is necessary to allow for separate effects from the relative price of housing and the real tax-adjusted mortgage interest rate. We also incorporat a measure to allow for the continual rise in owner-occupancy throughout the sample period. Under the assumption that house prices are market-clearing, the long-run demand curve can be recovered by inverting the long-run solution to an estimated house price equation. The resulting expression has the form:

ln(HD) = ln(C) + ln(O) - 0.553 ln(PH/PC) - 0.0036 [RM - 400*[Delta]ln(PH)] (2)

Here HD, O, PH and PC denote the real value of personal sector housing assets, owner-occupancy, average house prices and the consumers' expenditure deflator. The final term in (2) is a tax-adjusted real mortgage rate. Although the direct interest-rate effect in (2) appears small, it should be remembered that additional effects will emerege through the consumption equation (1).

Second, the model of the flow of housing supply (that is, new housebuilding) is explicitly based on the profit-maximising decisions of housebuilders. The long-run solution to the equation for housing starts, see (3), contains effects from expected profitability and expected house price inflation, given the time expected to complete construction (see Tsoukis and Westaway (1994) for a full description).

ln(HS) = 0.69 ln(EPH/CC) - 0.0079 [RC - 400 *[Delta]ln(EPH)] (3)

Here, HS denotes private sector housing starts, CC construction costs, RC a corporate tax adjusted interest rate. EPH denotes 'expected' future (new) house prices.

Personal sector investment in new dwellings is only recorded once a completed dwelling has been sold. We condition such investment on housing completions, allowing for an additional negative influence from tax-adjusted real mortgage rates to capture the opportunity cost of house purchase. Home improvement expenditure, the other component of aggregate personal sector housing investment, grows in line with consumer durables' expenditure and housing turnover. A separate equation is adopted for this component of investment as it is behaviourally distinct from new housebuilding, being initiated from the demand side by the owner-occupier, rather than from the supply-side by the builder.

In practice, it is unlikely that house prices adjust sufficiently so as to continuously equalise the stock demand and supply of housing assets. Imbalances between the two give rise to short-run 'disequilibrium' in the housing market. Given the assumption of eventual market clearing, the extent of any disequilibrium determines both the direction and scale of the dynamic adjustmen of aggregate house prices. Chart 4 below illustrates the derived series for 'housing disequilibrium'. It can be seen that periods when prospective demand exceeds supply have coincided with periods of rapid house price inflation. The present balance of pressures within the housing market now points to renewed price rises in the medium term, reflecting both the recent reductions in real interest rates and the rise in consumers' expenditure.

5. Within-sample dynamic simulations

This section summarises the results of a number of within-sample dynamic tracking exercises undertaken on the new system of equations for the personal sector. We examine the extent to which our proposed model is able to capture th behaviour of the housing market and consumers' expenditure over the period from 1985 to 1992, taking the historical values of non-property income, interest rates and financial asset prices as given. The property income of the personal sector was treated as endogenous, given the endogeneity of mortgage and credit demand, with consequential effects on mortgage interest tax relief, personal disposable income, the net acquisition of financial assets by the personal sector and net financial wealth.

In interpreting the results from these exercises it is important to bear in min that the dynamic simulations are merely a diagnostic device designed to illustrate the dynamic properties of a particular subsection of the model, rather than a means of formally evaluating whether the model is misspecified. Pagan (1989), in discussing the use of dynamic simulation procedures for evaluating (linear) models, demonstrates that they merely provide an alternativ means of summarising the information available from the individual equation residuals themselves.

Two variants were undertaken. The first uses the default model settings for the personal sector equations. The second variant was undertaken with consumer credit and equity withdrawal held at their historical values. The outcomes from these runs are summarised in Table 1. In the initial simulation (denoted S1) it remains the case that both the strength of consumer demand in the mid-1980s (particularly in 1986) and the depth and duration of the last recession are under-estimated. Growth of over 1 1/2 per cent is produced for 1992, a year in which consumption remained at its 1991 level. Even so, developments in the housing market are tracked rather well, particularly at the height of the boom in 1987/88. One surprising feature is that the model suggests that house prices might have been expected to fall even further than they did during the last recession.

The second variant simulation held consumer credit and equity withdrawal (hereafter 'borrowing') at their actual levels in order to assess whether the consumption 'prediction' errors in the recession arise from a failure to captur adequately the dynamics of the credit cycle rather than from any fundamental misspecification TABULAR DATA OMITTED of the consumption function itself. Table 1 shows that borrowing is underestimated by some [pounds]5 1/2 billion per annu in 1987/88 and overestimated by some [pounds]12 billion in 1992, when it declined to a lower level than at any time since 1980.

Allowing for the behaviour of credit removes the considerable over-estimate of expenditure in the recession. This suggests that while the disposable funds consumption function provides a satisfactory ex-post explanation of recent consumer behaviour, it would have been difficult to successfully predict the full extent of the recession ex-ante, because of the difficulties in anticipating the unexpectedly sharp decline in borrowing.

Finally, it is possible to estimate the impact of the pre-announcement in the 1988 Budget of restrictions on mortgage interest tax relief from the autumn of that year, since explicit allowance for this is made in the estimated house price equation. Re-running the dynamic simulation without this term suggests that the measure may have increased calendar year house price inflation by 4 1/2-5 per cent in 1988, and contributed significantly to the precipitate slowdown in 1989 and 1990. Further details are given in Pain and Westaway (1994).

6. The impact of consumer borrowing and equity withdrawal

It seems clear from the work presented so far that changes in the levels of consumer credit and equity withdrawal have had important effects on the macroeconomy. In particular, they are likely to have influenced the scale of both the late-1980s boom and the subsequent recession. In this section we repor the outcome of some 'counterfactual' simulations designed to analyse the contribution that the size and volatility of these particular components of total personal sector borrowing has made to recent macroeconomic developments.

Two sets of simulations are reported. The first uses the estimated set of equations for personal sector behaviour, allowing for the endogeneity of property income and wealth accumulation, to ask what would have been the outcom for the economy if borrowing had simply risen in line with (ex-ante) nominal personal disposable income from the beginning of 1985. The second counterfactua exercise seeks to quantify the extent to which the cyclical behaviour of borrowing can be attributed to changes in the loan-value ratio for first time buyers and speculative activity in the housing market. Both counterfactual simulations allow for the second-round impact of changes in consumption and the housing market on the wider economy.

All the simulations reported here make use of 'adaptive' expectations models with expectations formed solely on the basis of past events. The full Institute macroeconometric model incorporates explicit anticipatory behaviour in company sector factor demands, asset prices, wage and price setting behaviour as well a in the personal sector model described above. An important issue in any counterfactual exercise is the extent to which knowledge of the future course o key exogenous variables, such as world demand and fiscal and monetary policy, i allowed to affect behaviour in the present day. Use of 'model-consistent' expectations could well give misleading estimates, as it ensures that the actions of policymakers and developments within the wider world economy were perfectly anticipated in the latter half of the 1980s and the early part of thi decade. Two possible (computationally demanding) alternatives to the use of adaptive expectation models are the stacked solution and 'learning' techniques proposed by Wallis and Whitley (1992) and Westaway (1992).

A further issue of importance concerns the assumptions made about the policy regime in place in the counterfactual scenario. Although it may be of interest to assess the importance of particular changes under a given stance for fiscal and monetary policy, it is unlikely that the observed stance was fully independent of economic developments. Thus we have augmented the standard Institute model of the UK economy with a reaction function for short-term interest rates, estimated over the period from 1978 to 1987. Domestic interest rates are found to be related to foreign interest rates, changes in the exchang rate, retail price inflation, monetary growth and capacity utilisation. Full details are given in an appendix. Related results are contained in Gurney (1988 and Britton (1991). The exchange rate was endogenised using a modified open arbitrage condition, with the current period exchange rate related to the exchange rate and the differential between UK and world three month interest rates in the previous period. Tax rates and the volume of non-cyclical public expenditure were held at base levels.

A final caveat stems from the nature of the counterfactual analysis. In practic individual model equations all have residuals, reflecting the difference betwee past events and the model prediction of them. It is assumed that these elements in the past that the model cannot explain remain unchanged in the counterfactua run.

The results from the initial counterfactual with borrowing held at 'trend' levels are summarised in Table 2, with Charts 5 and 6 reporting the consumers' expenditure and house price effects in greater detail. Overall, the counterfactual scenario reduces the volatility of the economic cycle over this period, with the level of consumers' expenditure reduced by over 2 per cent at the height of the boom in 1988, but higher than before after 1990. House prices are reduced by some 5 1/2 per cent in 1988, but by 1993 are nearly 8 per cent above their actual value. The implied changes in real house prices help to change the profile of housing investment, with investment reduced in the late-1980s and raised between 1991 and 1993. The changes in GDP are smaller tha the changes in the components of domestic demand due to offsetting changes in net trade volumes.

The charts illustrate that the endogeneity of monetary policy helps to reduce the size of the counterfactual responses, with interest rates reduced a little below their base values in 1987 to 1990, and raised above their base values fro 1991. By 1993 UK interest rates in the counterfactual are some 2 percentage points above their actual values at this time. The initial reduction in domesti interest rates helps to generate an exchange-rate appreciation in order to equalise the returns on assets denominated in sterling and foreign currencies. The extent of the appreciation is modified after 1990 as UK interest rates rise above their base levels.(3)

Table 2. The impact of constant borrowing-income ratios 1985-93

all figures change from base levels

 1985 1988 1991 1993

GDP (%) -0.15 -1.30 0.89 2.23
Consumers' expenditure (%) -0.29 -2.10 1.78 4.02
Housing investment (%) -0.43 -4.67 2.41 5.23
Imports (%) -0.25 -1.62 2.15 3.40
Consumer prices (%) 0.00 -0.84 -2.03 -1.92
House prices (%) -0.34 -5.44 3.63 7.83
Employment (000s) -7 -176 -63 185
Current account/GDP (% pts) 0.08 0.60 -0.54 -1.10
PSBR/GDP (% pts) 0.05 0.40 -0.07 -0.55
Interest rates (% pts.) -0.08 -1.41 0.25 2.00
Exchange rate (%) 0.03 1.75 3.92 1.81

Memorandum item: GDP growth (% pts) -0.15 -0.68 1.11 1.13

Whilst the simulation does not entirely remove the recession, it does reduce it length. In the variant, GDP at constant prices falls by 1 1/2 per cent in 1991, but rises by nearly 3/4 per cent in 1992 (compared to actual falls of 2 1/4 and 1/2 per cent respectively). By the end of 1993, GDP in the UK had finally returned to its pre-recession peak. In the counterfactual the level of GDP at this date is some 1 1/2 per cent above the pre-recession peak, and some 2 1/4 per cent above the actual level of output at this time. Correspondingly, employment is above its base level by some 185,000.

The counterfactual shows marked changes in the profile of the levels of both mortgage equity withdrawal and consumer credit. Equity withdrawal is reduced by some [pounds]5 1/2 billion in 1987 and [pounds]7 1/2 billion in 1988 and raised by over [pounds]13 billion in both 1992 and 1993. Nominal new consumer credit i reduced by [pounds]2 1/2 billion in 1988, but raised by some [pounds]6 billion in 1992. In the absence of a model that can fully account for all of the observed variation in borrowing over this period, it is difficult to say to wha extent these changes can be attributed to 'financial liberalisation', although the absence of many regulatory controls undoubtedly facilitated the observed rise in personal sector debt-income ratios in the latter half of the decade.

A further set of counterfactual simulations were run in order to ascertain the extent to which fluctuations in the level of property transactions and the loan-value ratio could account for the changes in the level of borrowing impose in the first simulation. Two changes were made. First, the average loan-value ratio for first-time buyers was held at 80 per cent, roughly in line with the value immediately prior to deregulation. Second, the level of property transactions was constrained to merely grow in line with trends in owner-occupancy since the beginning of 1985. The difference between the resulting level of transactions and the historical data gives an approximate indication of the extent of speculative demand pressures within the housing market. The change reduces the level of transactions by some 300,000 (14 per cent) in 1988, but raises the level by some 800,000 in 1993.

Chart 7 illustrates the differences in the level of borrowing in the two counterfactual scenarios we consider (simulation value less the base value). Th changes to the average loan-value ratio and the level of property transactions account for around half of the changes in consumer credit and equity withdrawal imposed in the initial simulation. The resulting economic impact is summarised in Table 3. As might be expected the effects are around half the size of those in Table 2, although both simulations reduce the impact of the recession in 199 by a similar extent. This partially reflects the severity of the decline in the number of property transactions at this time and the consequential impact this had on home improvement expenditure.

It is of some interest to note that whilst both simulations imply a fall in rea house prices after 1988 neither implies a fall in nominal house prices. To the extent that factors such as the rise in the level of housing repossessions and the prevalence of negative equity are related to unexpected changes in nominal (rather than real) house prices, the steps taken to liberalise credit markets i the 1980s can be said to have helped to accentuate the severity of the most recent recession.

Conclusions

In this note, we have described the new system of equations for personal sector behaviour incorporated in the NIESR UK model. We have assessed the extent to which our system of equations is capable of explaining the boom of the late-1980s and the recession of the early-1990s. The roles of consumer credit and mortgage equity withdrawal are particularly important in this regard since direct effects from these are included in our consumption equation whereas thei determinants are implicit, or possibly absent, in conventional specifications. In explaining the boom in consumer spending in the late-1980s, we find that our consumption equation implies a single equation ex post forecasting error in consumption growth of 3/4 per cent in 1988, even allowing for use of the fitted values from our credit equations instead of the actual values. This error represents a considerable improvement on the errors of more than 3 per cent which were made at the time by most macroeconomic forecasters. These errors are similar to those now obtained if a more conventional specification for the consumption function is adopted.

Table 3. The impact of changes in the loan-value ratio and housing turnover

all figures change from base levels

 1985 1988 1991 1993

GDP (%) -0.11 -0.61 0.86 0.89
Consumers' expenditure (%) -0.19 -0.94 1.37 1.25
Housing investment (%) -0.56 -2.71 7.29 8.05
Imports (%) -0.18 -0.81 1.54 0.91
Consumer prices (%) 0.00 -0.33 -0.62 -0.18
House prices (%) -0.56 -2.38 4.54 1.91
Employment (000s) -5 -75 39 141
Current account/GDP (% pts.) 0.06 0.47 -0.47 -0.68
PSBR/GDP (% pts.) 0.03 0.19 -0.22 -0.29
Interest rates (% pts.) -0.05 -0.63 0.75 1.10
Exchange rate (%) 0.01 0.81 1.23 -0.62

Memorandum item: GDP growth (% pts) -0.12 -0.45 0.73 0.30

When explaining the depth of the recession in the early-1990s, however, our specification which conditions on the flow of credit provides a significantly better fit of the consumption equation compared to more conventional models, capturing all but 1/4 per cent of a trough in consumption growth in 1991 compared to an error of 1 1/4 per cent in the predictions from a conventional model and errors of over 2 per cent made in many forecasts at the time. This ability to explain the depth of the recession is somewhat reduced if credit is allowed to be determined by its behavioural equations, with the consumption forecast error increasing to just over 1 per cent in that year. This finding ha important implications for the modelling of personal sector behaviour. If we believe that the determinants of credit and mortgage equity withdrawal can simply be incorporated in the consumption function itself, then there is no particular advantage to conditioning on credit. On the other hand, if we believe, as our work suggests, that the determinants of these credit flows may be rather difficult to capture using conventional variables (that is, incomes, interest rates and so on), then it may be useful to separate out these influences and attempt to model credit explicitly.

Our simulation results illustrate the extent to which the rapid build-up of deb in the aftermath of financial deregulation can account for much of the observed instability in the UK economy since the middle of the 1980s. The relaxation of formal controls over lending resulted in households (and firms) acquiring a deb burden that was subsequently found to be excessive. The legacy of the recession has resulted in households adopting a more cautious attitude to debt, with bank and building societies imposing tighter financial criteria for prudential reasons. Even so, it remains possible that the experience of the 1980s could be repeated if people become more confident about their future prospects after a sustained period of economic growth. An assessment of the future growth of credit is thus an important part of any judgement as to whether the economy can proceed smoothly along a path of non-inflationary growth.

NOTES

(1) It is of interest to compare the experience of the early-1980s with the las recession. The official forecast in the 1979 Autumn Statement projected growth of only 0.5 per cent in consumers' expenditure in 1980, following growth of 4 per cent in 1979. This projected slowdown was closely in line with the eventual outturn. By contrast, the forecast in the 1990 Autumn Statement projected consumption growth of 1.75 per cent in 1991, resulting in an eventual forecast error close to 4 per cent.

(2) Whether it is preferable to model the individual components of aggregate equity withdrawal or the aggregate level alone is ultimately an empirical issue Here, we adopt the latter course, partly out of convenience since quarterly dat cannot easily be constructed for the individual components. However we make use of the preliminary disaggregated results of Westaway (1994) in choosing the factors which we expect to influence aggregate equity withdrawal.

(3) Over the period from 1990Q4 to 1992Q3 the UK was a member of the ERM. It could be argued that the interest rate reaction function be overwritten during this period, since the level of German interest rates acted as an effective floor for UK interest rates. In practice, the values for domestic interest rate generated by the reaction function were always at or above German ones, and so we continued to use this model.

Appendix: An Interest-Rate Reaction Function

The counterfactual analysis described in Section 6 makes use of a reaction function for UK short-term interest rates. This was estimated over a sample fro 1978 to 1987, during which time the prospects for inflation came to have a greater influence on the stance of monetary policy. The actual equation used wa derived using an instrumental variable technique due to the presence of current dated terms in the effective exchange rate.

TABULAR DATA OMITTED

Sample period:1978Q1-1987Q4; [Mathematical Expression Omitted]; Standard error 0.80; LM(4)=7.1; RESET(1)=0.91; NORM(2)=0.45; HET(1)=0.02; SARGAN(7)=7.43

Additional instruments:

[Delta]ln[(EFFRAT).sub.-1], [Delta]ln[(EFFRAT).sub.-2], [Delta]ln[(EFFRAT).sub.-3], [Delta]ln[(EFFRAT).sub.-4], [Delta]R[W.sub.-1], [Delta]R[B.sub.-1], [Delta]ln[(WPO).sub.-2], CB[Y.sub.-2]

Parameter stability 1988Q1-1990Q3: SALKEVER(11) = 10.33

Parameter stability 1988Q1-1993Q4: SALKEVER(24) = 28.64

Variable Definitions:-

EFFRAT = sterling effective exchange rate (1985 = 100)

RB = UK base rates

RWT = weighted world three-month interest rates

RINF = annual rate of retail price inflation

M0 = notes and coin in circulation

CUMF = capacity utilisation in manufacturing

WPO = world oil prices ($)

CBY = UK current account / GDP ratio

The post-sample parameter stability tests indicate that the parameters are (jointly) stable over the period to the end of 1993. However, the predictions are noticeably poorer during the last two years of available data, with interes rates being underpredicted in 1992 and over predicted in 1993.

REFERENCES

Blake AP and Westaway PF (1993), 'Why have UK consumption functions broken down?', National Institute Discussion Paper No. 38.

Blake AP and Westaway PF (1994), 'Small samples and structural change: a simulation study of consumption', National Institute Discussion Paper No. 63.

Britton A (1991), Macroeconomic Policy in Britain 1974-87, Cambridge University Press.

Britton A and Pain N (1992), Economic Forecasting in Britain, National Institut Report Series No. 4.

Carruth A and Henley A (1990), 'The housing market and consumer spending', Fiscal Studies, No. 11, pp 27-38.

Carruth A and Henley A (1992), 'Consumer durables spending and housing market activity', Scottish Journal of Political Economy, No. 39, pp261-271.

Cuthbertson K (1980), 'The determination of expenditure on consumer durables', National Institute Economic Review, No. 94, pp 62-72.

Darby J and Ireland J (1993), 'Consumption, forward looking behaviour and financial deregulation', presented to ESRC Macroeconomic Modelling Bureau Conference, Warwick University, July 1993.

Dicks MJ (1990), 'A simple model of the housing market', Bank of England Discussion Paper No. 49.

Ermisch J and Westaway PF (1990), 'The dynamics of aggregate consumption in an open economy life cycle model', National Institute Discussion Paper No. 190, Scottish Journal of Political Economy, forthcoming (1994).

Gurney A (1988), 'The exchange rate, interest rates and the current balance in forward looking model', National Institute Economic Review, No. 125, pp 40-46.

Holmans AE (1991), 'Estimates of housing equity withdrawal by owner occupiers i the United Kingdom, 1970 to 1990', Government Economic Service Working Paper No 116.

Holmes MJ (1993), 'Housing equity withdrawal and the average propensity to consume', Applied Economics, No. 25, 1315-1322.

Kilpatrick A (1991), 'The performance of broad money', Treasury Bulletin, Autum 1991.

Meen GP (1990), 'The measurement of mortgage market constraints and the implications for econometric modelling of UK house prices', Oxford Bulletin of Economics and Statistics, No. 52, pp 1-23.

Melliss C, Meen G, Pain N and Whittaker R (1989), 'The new Treasury model project', Government Economic Service Working Paper, No. 106.

Miles D (1992a), 'Housing and the wider economy in the short and long run', National Institute Economic Review, No. 139, pp 64-77.

Miles D (1992b), 'Housing markets, consumption and financial liberalisation in the major economies', European Economic Review, No. 36, pp 1093-1136.

Muellbauer J and Murphy A (1989), 'Why has UK personal saving collapsed?', Credit Suisse First Boston Research Paper, July 1989.

NIESR (1994), National Institute Domestic Econometric Model Manual, National Institute of Economic and Social Research, May 1994.

Pagan A (1989), 'On the role of simulation in the statistical evaluation of econometric models', Journal of Econometrics, No. 40, pp 125-140.

Pain N (1994), 'The Economic Impact of the 1993 Budgets', National Institute Economic Review, No. 147.

Pain N and Westaway PF (1994), 'Housing, consumption and borrowing: an econometric model of personal sector behaviour', GES Working Paper forthcoming.

Patterson KD (1993), 'Consumers' expenditure on nondurables and services and housing equity withdrawal in the United Kingdom', University of Reading Discussion Papers in Quantitative Economics and Computing, Series E vol II, No. 8.

Sargent JR (1991), 'Deregulation, debt and downturn in the UK economy', Nationa Institute Economic Review, No. 137, pp 75-88.

Tsoukis C and Westaway PF (1994), 'A forward-looking model of housing construction in the UK', Economic Modelling, No. 11, pp 266-279.

Wallis KF and Whitley JD (1992), 'Counterfactual analysis with macroeconometric models: the UK economy 1979-84', ESRC Macroeconomic Modelling Bureau Discussion Paper No. 29.

Westaway PF (1989), 'Modelling consumers' expenditure in the UK: a disposable funds approach', presented at European Economic Association Conference, Lisbon, 1990.

Westaway PF (1992), 'A simulation model of consumer spending and housing demand', National Institute Discussion Paper No. 15.

Westaway PF (1993), 'A micro to macro model of personal sector behaviour', pape presented to HM Treasury Academic Panel, AP(93)8.

Westaway PF (1994), 'Mortgage equity withdrawal: causes and consequences', National Institute Discussion Paper No. 59.