financial

1 23

Social Indicators Research An International and Interdisciplinary Journal for Quality-of-Life Measurement ISSN 0303-8300 Soc Indic Res DOI 10.1007/s11205-016-1409-z

Tenure and Spending Within UK Households at the End of the Recent Recession

Jacinta C. Nwachukwu

1 23

Your article is protected by copyright and all

rights are held exclusively by Springer Science

+Business Media Dordrecht. This e-offprint

is for personal use only and shall not be self-

archived in electronic repositories. If you wish

to self-archive your article, please use the

accepted manuscript version for posting on

your own website. You may further deposit

the accepted manuscript version in any

repository, provided it is only made publicly

available 12 months after official publication

or later and provided acknowledgement is

given to the original source of publication

and a link is inserted to the published article

on Springer's website. The link must be

accompanied by the following text: "The final

publication is available at link.springer.com”.

Tenure and Spending Within UK Households at the End of the Recent Recession

Jacinta C. Nwachukwu1

Accepted: 5 July 2016 � Springer Science+Business Media Dordrecht 2016

Abstract Housing costs form a substantial share of aggregate demand in the UK. This study examines the distribution of total expenditure-to-income by homeownership status at

the end of the recent recession in 2010. Multivariate quantile regressions uncover four

important points. First, owner-occupiers in England have considerably higher mean

spending ratios than their peers in other parts of the UK; an indication of their wealthier

status. Second, the average spending ratio for residential-occupiers in all UK regions, with

the exception of Northern Ireland, is significantly higher than the mean ratio for tenants in

both private and public properties. In this last region, the spending rate for private tenants

is more prominent. Third, the disparity in the expenditure ratio between owner-occupiers

and tenants is significantly more pronounced in England. Fourth, renters in public housing

in Scotland and Wales have much higher spending ratios than their counterparts in private

properties, reflecting a greater overall social security provided by the devolved government

there. Policy implications allied with heterogeneity in the consumption effect of housing

wealth across the different homeownership cohorts is discussed.

Keywords Household consumption � Housing tenure � Economic recession � Quantile regression

JEL Classification O18 � O50 � C21

1 Introduction

There is a general perception that a corollary of the financial crises of 2007–2009, and the

subsequent economic recession, was a considerable fall in asset value, in particular houses in

the UK. For example, the Nationwide House Price Index (2011) lost around 8 % of its value

& Jacinta C. Nwachukwu [email protected]

1 School of Economics, Finance and Accounting, Faculty of Business and Law, Coventry University, Priory Street, Coventry CV1 5FB, UK

123

Soc Indic Res DOI 10.1007/s11205-016-1409-z

Author's personal copy

between the first quarter of 2007 and the last quarter of 2009, leading to a negative equity for

many home owners. Basic economic theory suggests such shocks on asset prices are likely to

lead to a cut in consumer spending in favour of precautionary savings, especially if house-

holds expect the recent turmoil that has been experienced in the global economy to persist for

the foreseeable future (Berry and Williams 2009; Crossley and O’Dea 2010).

This observation has prompted much discussion in the contemporary literature and

provided the justification for the substantial stimulus delivered by the Bank of England’s

expansionary monetary policy and quantitative easing (Scholz et al. 2006; Khoman and

Weale 2008). In addition, Crossley et al. (2012) noted that a number of economic recovery

initiatives by the UK Government over the years have centred on how to help people buy

and occupy new or existing houses. The latest example of such housing purchase options is

the ‘‘Help to Buy Equity Loans and Mortgage Guarantees’’ introduced in March 2013 to

assist people to acquire residences priced up to £600,000 with as little as a 5 % deposit. All

these policies are motivated, at least in part, by the concern that if their wealth is con-

sidered to be too low by credit-constrained home owners, then they will consume less of all

private commodities and/or even save more of their existing capital with a fall in economic

activity.

This paper contributes to the debate on housing policy by answering two key questions.

They are: (i) how much did those who occupied their properties consume out of their

household earnings compared with those who rented either from private landlords or public

authorities at the end of the recent economic recession in 2010 and (ii) did aggregate

consumption relative to income of home owner-occupiers vary significantly across the

different UK regions at this crisis end year.

This study attempts to deal with these concerns in two ways. First, it examines the

distribution of aggregate average household expenditure per week against income (here-

after referred to as the expenditure ratio) for the different types of housing occupancy

across the four regions of the UK—Northern Ireland, Wales, Scotland and England. Se-

cond, it investigates the major determinants of UK household spending on all products and

services relative to income in 2010. We focus on the outlay profile of property owner-

occupiers in the belief that an examination of how UK consumers responded to recession is

best captured by the impact of unexpected shocks to the market value of houses with

related spending adjustment by residential owners compared with tenants in both private

and public sectors.

This paper is organised as follows: Sect. 1 summarizes the theoretical and empirical

literature on household consumption. We pay particular attention to those studies which

emphasise the role of housing wealth in predicting fluctuations in consumption activity.

Section 2 describes the data. Section 3 specifies the empirical models. Section 4 outlines

the results of our quantile regressions. The concluding section recommends policy to

support spending among home owner-occupiers in the UK.

2 Literature Review

The purpose of this section is discussed under two headings. The first summarizes the

theory of household consumption and savings. The second reviews the findings of previous

empirical studies which considered the factors that are likely to influence the aggregate

spending decisions of families, in particular, the effect of their housing wealth and home

ownership position.

J. C. Nwachukwu

123

Author's personal copy

2.1 Household Consumption and Savings Theory

The theoretical reasons why household spending and saving activities may vary is usually

based on the following identity equation (Disney et al. 2002; Sinai and Souleles 2005;

Berry and Williams 2009; Campbell and Cocco 2007)

St þ NDt ¼ NFAt þ NHt ð1Þ

where St is household savings representing the difference between current income and

consumption. The symbol NDt is the net acquisition of debt which is measured as new

loans minus repayment of principal on existing debt. Therefore, the sum of St þ NDt captures the totality of funds raised by households in a given year t. The right hand side of

the equation gives the sum of the net assets accumulated by households. For example, the

term NFAtð Þ is net financial assets measured as the purchase minus sale of financial assets. The symbol NHtð Þ is the accumulation of houses defined as the acquisition minus repay- ment of mortgage principals. Taken together, we may infer from the identity equation that

the more households consume out of their current incomes, the more likely they are to

incur new debt. Otherwise, households which dissave and which are unable to raise net

loans would have to seek funds from other sources, including the sale of existing financial

and housing assets, in order to support current purchasing power. Thaler (1990) and Disney

et al. (2002) remarked that the magnitude of the propensity to consume out of housing

versus financial wealth may depend on the economic circumstances and age of a household

reference person.

Generally, four key broad theories are used to explain how household consumption

behaviour varies over time in order to re-establish equilibrium in the identity equation

above. They may be summarised as follows:

(i) The theory of Ricardian equivalence which dates back to a study by Ricardo in 1820

is among the earliest published hypotheses on the determinants of aggregate consumption

and savings. Subsequent theoretical arguments by Modigliani (1961), Diamond (1965),

Barro (1974), O’Driscoll (1977), Tanner (1978), Blanchard (1985), Feldstein (1986) and

Seater (1993) have clarified the assumptions upon which the conventional Ricardian

propositions depend in both the short and long-term. Under the Ricardian hypothesis,

deficit-finance tax cuts and government borrowing will exert no expansionary effect on

household spending. The reasoning is that rational and farsighted taxpayers will react to

such declines in government savings by paying off outstanding loans, acquiring new assets

and/or by accumulating bank savings deposits and the equity in their houses. Such is in

anticipation that expansionary fiscal policies merely postpone higher tax collection by the

authorities in the future. The implication is that consumers would be indifferent of the scale

and timing of taxes and government purchases and hence should not alter their spending

decisions. However, a number of authors including Modigliani and Sterling (1976), Kot-

likoff and Spivak (1981), Carmichael (1982), Abel (1985, 1986), Kotlikoff (1988),

Bernheim and Bagwell (1986) and Bernheim (1987) noted that the neutrality proposition of

Ricardian equivalence depends upon a set of implausible assumptions, including an effi-

cient capital market and the fact that a deferment of taxes does not lead to a re-distribution

of resources within generations.

(ii) The permanent-income hypothesis was originally introduced by Duesenberry (1949)

and Friedman (1957). They hypothesized that household income comprised permanent and

transitory components. The permanent income element reflects the effect of fundamental

factors such as the training, personality, occupation, status and location of employment

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

which affect the market value of household wealth. The transitory components are likely to

be considered by consumers as fortuitous occurrences such as rare illness, loss of job,

unexpected inheritances, windfalls or losses arising from changes in asset prices. Hall

(1978) reported that the implication of this transitory element is that consumption is likely

to follow a random walk. This means that the fraction of permanent income relative to total

income is the only factor powerful enough to systematically alter household spending

activities. Hence, families with current disposable income which is higher than that

anticipated in the future for their tax bracket as whole would save more in order to

compensate for the expected decline in their permanent income proportion, regardless of

favourable transient effects. Further evidence on the relative importance of permanent and

transitory income percentages for the smoothness of consumption was provided by

Modigliani (1966), Leland (1968), Mayer (1972), Deaton (1986), Hall (1981), Mankiw

(1981), Shapiro (1983), Mankiw and Shapiro (1985), Campbell (1987), Blundell (1988),

Attanasio and Weber (1994) and Carroll (2001). These studies concluded that while the

evidence on the permanent income hypothesis is generally favourable, the variability in

consumption with related precautionary savings appears to be smaller than predicted by the

theory. This indicates that households attempt to maintain a constant consumption-income

ratio. They attributed this failure to the fact that the traditional permanent income

hypothesis model is not robust to variable real interest rates and the presence of borrowing

and saving constraints. Under these extensions, expected consumption should fluctuate at a

rate proportional to the real rate of return and the degree of restrictions on the utility

function. An expected increase in the real rate of return should persuade families, par-

ticularly home owners with mortgages to pay off, to postpone present current spending.

What is more, prospective rises in future real interest rates would discourage expenditure,

in particular among retired people with investment in fixed assets such as bonds and

pension funds. Then too, interest charges on loans normally depend on the creditworthiness

of the borrower. Thus, individuals with a poor credit history, for example the unemployed,

face much higher borrowing costs and are more likely to be denied access to bank loans.

Their consumption spending will thus be transitorily curtailed.

(iii) Financial and housing wealth is related to movements in personal wealth and

consumption therefrom. Skinner (1989) and Millard and Power (2004) theorised that a rise

in the price of assets, including houses, could mean that individuals who hold them are

likely to raise their transaction and speculative demand for money. They proposed a

positive effect of housing wealth on consumption across households. Besides, Benito et al.

(2006) remarked that, a rise in the price of houses could be evidence that people who own

their own properties have more collateral against which to borrow, even when there are

housing inheritance motives. If credit becomes cheaper for them, then their spending may

be higher (Elliot 1980; Miles 1992, 1993, 1997; Bosworth et al. 1991; Attanasio and Weber

2010). Alternatively, Gale and Sabelhaus (1999), Poterba (2000) and Dynan and Maki

(2001) reported that an increase in house prices could force borrowers, especially would-be

first-time buyers, to accumulate higher deposit capital in financial assets such as bank

accounts, bonds and shares increases. Thaler (1990) commented that such active financial

savings are normally mentally designated as ‘‘non-fungible’’ accounts as a form of self-

control mechanism. This means that home owners could react differently to changes in

their realised gains in financial and housing wealth. Nevertheless, Lettau and Ludvigson

(2004), Edison and Slok (2001), Case et al. (2005) suggested that in line with the per-

manent income hypothesis, unexpected windfalls in financial and housing wealth must be

perceived as long-lasting to affect individual spending plans. Other studies which have

investigated the possible independent roles of both financial and housing wealth on

J. C. Nwachukwu

123

Author's personal copy

consumption include, Belsky and Prakken (2004), Carroll (2004), Chen (2006), Dvornak

and Kohler (2003), Campbell and Cocco (2007) and Bostic et al. (2009).

(iv) Demographic factors relating to education, marital status, gender and age are

captured by the life-cycle theory pioneered especially by Modigliani (1963, 1964, 1966),

Modigliani and Ando (1957) and Brumberg (1956). The basic life-cycle hypothesis

deviates from the previous theoretical models by assuming that consumption decisions of

households at each point in time do not depend solely on the basis of their tax proposals,

current income or financial and housing wealth. Rather households in planning their

consumption must take account of expected changes in their future life circumstances and

past experience. In particular, it is proposed that because individuals can forestall that their

incomes will fall considerably when they retire, they save when younger and dissave after

retirement. The outcome is a hump-shaped profile of spending over a person’s life-time.

The expenditure-to income ratios are expected to rise when people are young; aged

between 20 and 30 years. This population cohort has relatively low income and is more

likely to borrow against its anticipated higher future earnings in order to meet current

demands for schooling, marriage and child birth. But as it moves into middle age, it tends

to cut its spending ratio in favour of savings for retirement. The expenditure ratio for the

middle age group is predicted to peak between the ages of 40 and 60 years. As people

retire, their savings are run down to support spending on food, heating and lighting, health

and care assistance. But Miles (1997) and Danziger et al. (1982) remarked that the basic

life cycle theory is inconsistent with conditions where wealth fails to decline rapidly after

retirement due to government and intergenerational transfers to the elderly population.

Thus, conventional life-cycle hypothesis may overstate the true magnitude of dissaving for

the elderly than for the non-elderly population.

2.2 Empirical Literature

Empirical investigation of the patterns of household expenditure dates back to the work by

Ernst Engle in the mid-nineteenth century. The conjectures developed by Engle were

popularised by many writers, including Houthakker (1952, 1957), Prais (1952), Aitchison

and Brown (1954), Stigler (1954), Hirsch (1976) and Scitovsky (1976), in an attempt to

explain the nature of income-expenditure relationships in the twentieth century. These

authors argued that goods and services relating to basic and higher consumer needs display

distinctive income elasticities of demand. Such conclusions motivated Blundell et al.

(1993) to propose that empirical studies should be based on microdata on consumer

demand for singular products such as alcohol, clothing, energy and education. Alterna-

tively, Capps and Love (1983), Härdle et al. (1991), Manning et al. (1995), Engel and

Kneip (1996), Koenker and Hallock (2001), Ronning and Schulze (2004), and Caglayn and

Astar (2012) recommended that empirical researchers employ statistical methods, such as

tobit and quantile regression models, which deal explicitly with the heterogeneity asso-

ciated with the different categories of goods, time, geographic locations and intensity of

consumption.

Surprisingly, very little empirical work has been done on the pattern of consumer

spending and the factors influencing such expenditure decisions in the UK. Notable ex-

ceptions include the study by Atkinson et al. (1990), Blundell et al. (1993), Anderson, et al.

(1994), Miller (1998), Pahl (1999, 2000), Nickell (2004), Lise and Seitz (2011), Van de

Ven (2011) and Crossley et al. (2012). For example, Van de Ven (2011) observed that

consumer spending in the UK responded strongly to factors which influence individual

expectations on current vis-à-vis future developments in credit availability, employment,

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

demography and financial wealth. Lise and Seitz (2011) concluded that around two-thirds

of the differences in consumption allocations on goods categories within households can be

explained by the disparity in the earnings and hours worked by husbands and wives.

Crossley et al. (2012) found that the impact of an economic downturn on consumer

spending has been deeper in the most recent recession which occurred in 2008–2009 than

in the previous two which happened in the early 1980s and 1990s. In particular, the young

have cut back expenditure more than the old as have mortgage holders compared to private

and public renters. By contrast, the effect of the recession has been similar across the high

and low education attainment groups, partly due to state benefit and the UK’s progressive

tax system.

There are few prominent studies which specifically assess the marginal propensity to

consume out of housing wealth by the owner-occupiers normally cited in the literature.

They include the research by Skinner (1989) and Engelhardt (1996) who found a positive

impact of house price shocks on household consumption in the United States. Similarly,

Muellbauer and Murphy (1997), Carruth and Henley (1990), Miles (1993, 1997) reported

an affirmative marginal propensity to consume from housing wealth for UK residents.

Disney et al. (2002) extended the methodology employed in these earlier studies by

examining the degree of asymmetric response of consumption to gains and falls in house

prices in the UK. They found greater responsiveness of consumption to house price gains

than falls for owners-occupiers with zero or negative equity values, especially among

elderly households who may be unwilling to move in order to release housing wealth.

Belsky and Prakken (2004), Carroll (2004) and Campbell and Cocco (2007) concluded that

appreciation in housing wealth generally results in increased consumption by younger

owner-occupiers who tend to be less cautious in spending those gains.

The review of literature in this section illustrates that the importance given to a robust

analysis of household expenditures by researchers and policymakers has risen considerably

in the past decade. The current paper adds to this debate by comparing the patterns of

consumer spending relative to the gross income of property owner-occupiers with those of

renters in the UK in the year 2010, soon after the most recent global crises. In contrast to

most prior research, we disaggregate the data on the expenditure ratio into total quantile in

order to test for differential house wealth estimates across the low and high spending

categories.

3 Data

This section gives a descriptive account of our data under the following headings: (1) the

trend in regional household expenditure ratios in the UK, (2) data distribution and (3)

housing tenure and household expenditure ratios across the different UK regions.

3.1 The Trend in Regional Household Expenditure Ratios in the UK

Table 1 contains the description of the dummy variables used in our classification of UK

households. Also, we provide in Table 2 some descriptive statistics on the average

expenditure ratio for our overall sample and for the following three sub-groups of housing

tenure: (i) house owner-occupiers, (ii) private tenants and (iii) public tenants. The overall

dataset contains the entire 5263 households which continuously kept a diary record of the

family’s daily spending on the thirteen categories of goods and services identified by the

J. C. Nwachukwu

123

Author's personal copy

designers of the Living Costs and Food (LCF) questionnaire in 2010. Figure 1 charts the

pattern of the ratio of aggregate weekly spending to gross income for households in

Northern Ireland, Wales, Scotland, England and for all the respondents in our study

sample. For comparative purposes, we have also included the average figures for the UK as

a whole in 2010 and for the period 2000–2010 obtained from the Office for National

Statistics (ONS, Family Spending, 2011). 1 Two key features stand out.

The first is that families in Wales, Northern Ireland and Scotland spent below our study

sample average of 124.14 % of gross income. What is more, the expenditure ratios for the

last two regions are more alike than the ratio for England. Such is presumably an indication

of the resemblance in poverty levels and the social welfare benefits provided by the

devolved regional assemblies in Northern Ireland and Scotland. By contrast, respondents in

Table 1 The definition of independent variables used in the analysis

Variable symbol Variable name Description

LCY Logarithm of expenditure ratio The natural logarithm of the ratio of total weekly expenditure by adults and children divided by gross nominal income.

TENURE Tenure type House owned = 1 House is privately or publically rented = 0

REGIONS Government office region England = 1 Other regions = 0

TENREGNS The product of tenure and regions Homeowners in England = 1 Others home occupancy types = 0. This comprises homeowners in other parts of the UK and renters in England and the other UK regions.

WAGE Main source of household income Earned income = 1 Other sources of income = 0

INTERNET Internet connection in household Household has internet connection = 1 Household has no internet connection = 0

EMPLOYMENT Economic position of household reference

Economically active = 1 Economically inactive = 0

CLASS Class of household reference person Higher managerial classes = 1 Other working classes = 0

GENDER Sex of household reference person Male = 1 Female = 0

HSIZE Number of persons in household Three persons or more = 1 Less the three persons = 0

ADULTS Number of adults in household Two adults or more = 1 One adult = 0

CHILDREN Number of children in household Children in the household = 1 No child in the household = 0

1 The figures from the ONS are weighted averages created using population data from the 1991-2001

Census. They are therefore not directly comparable to the equally weighted average spending ratios which underlie our analysis in this paper. Nevertheless, we have chosen to include the ONS figures in order to provide the reader with a benchmark on which to relate any discrepancy in the expenditure behaviour of our responding households in 2010.

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

Table 2 Weekly aggregate expenditure to income ratios of UK households (mean in the year 2010)

House owner- occupiers

Private tenants

Public tenants

All respondents

Government office regions

Northern Ireland (0) 98.33 100.32 93.86 98.03

Wales (1) 109.53 95.19 108.44 107.60

Scotland (2) 103.10 83.98 93.44 98.46

England (3) 140.90 118.13 88.63 128.74

Overall UK sample 134.85 113.88 90.21 124.14

1 Government office regions (regions)

a Other regions (category, 0) 104.29 90.31 97.29 101.11

Number of observations 594.00 116.00 166.00 876.00

Observations (% total sample) 16.52 15.26 18.32 16.65

b England (category, 1) 140.90 118.13 88.63 128.74

Number of observations 3002.00 644.00 740.00 4386.00

Observations (% total sample) 83.48 84.74 81.68 83.35

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) 36.61 27.81 -8.66 27.62

2 MAIN source of household income (Wage)

a Other sources of income (category, 0) 226.17 170.22 93.77 183.47

Number of observations 1413.00 274.00 632.00 2319.00

Observations (% total sample) 39.29 36.05 69.76 44.07

b Earned income (category, 1) 75.75 82.12 82.01 77.38

Number of observations 2183.00 486.00 274.00 2943.00

Observations (% total sample) 60.71 63.95 30.24 55.93

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) -150.42 -88.10 -11.77 -106.09

3 Internet connection in household (internet)

a No internet (category, 0) 132.62 101.42 84.31 112.88

Number of observations 750.00 180.00 446.00 1376.00

Observations (% total sample) 20.86 23.68 49.23 26.15

b Has internet connection (category, 1) 135.44 117.75 95.93 128.12

Number of observations 2846.00 580.00 460.00 3886.00

Observations (% total sample) 79.14 76.32 50.77 73.85

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) 2.82 16.33 11.62 15.24

4 Employment position of household reference person (employment)

a Unemployed or retired (category, 0) 130.17 182.15 91.26 123.93

Number of observations 1279.00 177.00 560.00 2016.00

J. C. Nwachukwu

123

Author's personal copy

Table 2 continued

House owner- occupiers

Private tenants

Public tenants

All respondents

Observations (% total sample) 35.57 23.29 61.81 38.31

b Full or part time employed (category, 1) 137.44 93.15 88.52 124.27

Number of observations 2317.00 583.00 346.00 3246.00

Observations (% total sample) 64.43 76.71 38.19 61.69

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) 7.26 -89.00 -2.73 0.34

5 Class of household reference person (class)

a Other working classes (category, 0) 164.98 130.70 91.25 143.23

Number of observations 2355.00 523.00 859.00 3737.00

Observations (% total sample) 65.49 68.82 94.81 71.02

b Higher managerial classes (category, 1) 77.69 76.77 71.26 77.35

Number of observations 1241.00 237.00 47.00 1525.00

Observations (% total sample) 34.51 31.18 5.19 28.98

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) -87.29 -53.93 -19.99 -65.88

6 Sex of household reference person (gender)

a Female (category, 0) 107.88 109.32 92.06 104.10

Number of observations 1210.00 337.00 526.00 2073.00

Observations (% total sample) 33.65 44.34 58.06 39.40

b Male (category, 1) 148.53 117.51 87.66 137.16

Number of observations 2386.00 423.00 380.00 3189.00

Observations (% total sample) 66.35 55.66 41.94 60.60

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) 40.65 8.19 -4.40 33.06

7 Household size, no of persons in household (HSIZE)

a Less the three persons (category, 0) 160.19 101.61 90.90 139.22

Number of observations 2343.00 506.00 623.00 3472.00

Observations (% total sample) 65.16 66.58 68.76 65.98

b Three person or more (category, 1) 87.48 138.32 88.69 94.88

Number of observations 1253.00 254.00 283.00 1790.00

Observations (% total sample) 34.84 33.42 31.24 34.02

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) -72.71 36.70 -2.21 -44.34

8 Number of adults in household (adults)

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

England spent almost 5 percentage points above the study sample average. Information

from the Office for National Statistics consistently indicates that English residents, on

average, pay more for housing together with fuel and power, household goods and services

including home improvement and insurance, health services including private medical

treatment and pharmaceutical items as well as recreation and cultural events. Besides,

English residents, especially those in London and the South East, are wealthier than

elsewhere in the UK. Statistics show that they have the highest spending amongst our

households on education and on luxury items like restaurants and hotels, communication,

transport and miscellaneous items which may include holiday overseas, legal protection

and personal services from domestic servants, exercise trainers and nannies or au pairs.

Second, residents in Wales and England spent more than their average weekly incomes

in 2010. Consequently, the mean spending ratio for our study sample of 124.14 % is

considerably higher than the population-weighted average figure of 66.99 and 70.77 %

reported by the ONS for the UK as a whole in 2010 and from 2000 to 2010 respectively.

Two propositions may be deduced from the identity equation in Sect. 1.1. They are: (i) the

Welsh and English inhabitants are increasing their net debt burden either by acquiring

further new loans and/or failing to repay interest or principal on existing loans in full and

on time and (ii) the majority of our Welsh and English dwellers could be drawing down

their assets, including equity in their houses, in order to maintain the level of consumption

for which they have become accustomed. We may therefore insinuate a positive correlation

Table 2 continued

House owner- occupiers

Private tenants

Public tenants

All respondents

a One adult (category, 0) 154.13 116.82 91.87 129.45

Number of observations 978.00 326.00 533.00 1837.00

Observations (% total sample) 27.20 42.89 58.83 34.91

b Two adults or more (category, 1) 127.65 111.67 87.84 121.29

Number of observations 2618.00 434.00 373.00 3425.00

Observations (% total sample) 72.80 57.11 41.17 65.09

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) -26.48 -5.15 -4.04 -8.16

9 Number of children in household (children)

a No children in the household (category, 0) 151.38 106.82 88.25 135.29

Number of observations 2597.00 496.00 588.00 3681.00

Observations (% total sample) 72.22 65.26 64.90 69.95

b Children in the household (category, 1) 91.89 127.14 93.84 98.17

Number of observations 999.00 264.00 318.00 1581.00

Observations (% total sample) 27.78 34.74 35.10 30.05

c All respondents 134.85 113.88 90.21 124.14

Number of observations 3596.00 760.00 906.00 5262.00

d Difference in the mean spending ratio

Item (b) minus item (a) -59.49 20.31 5.59 -37.13

J. C. Nwachukwu

123

Author's personal copy

between housing wealth and consumption rates. Further, we may propose that regional

heterogeneity in house prices have an important effect on household consumption rates

since our figures suggest that the impact of the wealth effect is highest for homeowners in

England, followed by Wales. This may be because housing is an asset that can be used as

collateral for a loan. Thus, the higher property prices in England in particular may have

allowed borrowing constrained homeowners to smooth consumption.

3.2 Data Distribution

Discussion here comprises (i) a brief description of the nature of our household-level data

provided by the Living Costs and Food Survey (LCF), (ii) a distribution of respondents

across regions and (iii) a distribution of respondents across homeownership status.

3.2.1 The Living Costs and Food Survey (LCF)

The Living Costs and Food Survey (LCF) provides the microdata used in our study. The

LCF which was formally known as the Expenditure and Food Survey from 2001 to 2008 or

the Family Expenditure Survey prior to 2001 was introduced in 2001/2002 by the Office

for National Statistics (ONS) and the Department for Environment, Food and Rural Affairs

(DEFRA). The aim was to collect information on the purchasing habit of private house-

holds and individuals aged 16 and above in the UK. The selection of households for an

LCF survey was based on a multi-stage stratified random sample design in order to

maintain the proportion of households in each region of the UK population.

Expenditure data for each household was garnered in two ways. First, through an

interview which was carried out once per household in the relevant financial year, resulting

in around 1750 households been interviewed in each quarter, so that in a typical financial

Fig. 1 The average for the weekly expenditure to gross income ratio in 2010

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

year there are potentially 7000 observations available. Second, over a two-week period, the

adult members of each household were asked to keep a diary of their consumption

expenditures on durable and non-durable goods and services. In addition the survey con-

tained a variety of other information, including the region where the household lived, gross

income, major source of income, economic status, demographics such as gender, age and

household composition, social class and homeownership status. The key advantages and

methodological limitations of the LCF Survey are detailed in many of the studies identified

in the empirical literature in Sect. 2.2, more specifically, Campbell and Cocco (2007),

Collis et al. (2010), Purshouse et al. (2010) and Meng et al. (2014).

3.2.2 The Distribution of the 5263 Respondent Households

The distribution of the 5263 respondent households in our study sample is depicted in

Fig. 2 across the four UK regions—England, Scotland, Wales and Northern Ireland.

It shows that our dataset for analysis is dominated by respondents from England.

Collectively families in England make up around 83 % of our sample and by inference of

the UK as a whole. This outcome is to be expected given that respondents are selected from

a stratum rather than from a universe of the entire UK households.

Further, Fig. 3 shows a frequency distribution of our respondents in a histogram with

data on the range of household spending ratios presented on the horizontal axis. About

three-quarters of respondents in our sample have average expenditure ratios under 100 %.

Interestingly 0.4 % of households reported an expenditure ratio in excess of 1000 %. Such

resulted in a distribution that is positively skewed with a long right tail. Indeed, a test for

Fig. 2 The distribution of respondents in the survey across regions

J. C. Nwachukwu

123

Author's personal copy

normality using the Jarque–Bera test statistic rejects the null hypothesis that our expen-

diture ratio series is normally distributed. This indicates that the raw data will have to be

transformed using common techniques, such as natural logarithms, in order to obtain the

appropriate functional form for an empirical modelling of the determinants of household

spending in the UK. The manner in which this data transformation is carried out for our

empirical model is discussed in the subsequent section.

Also, in Fig. 4, we exploit the information in our dataset by grouping cohorts in terms of

their homeownership status, regardless of the UK region where the household lives.

The data for the residential-owner occupiers dominates with more than two-thirds of the

respondents in our sample claiming to own the house in which they reside. This outcome is

to be expected given the prevalence of homeownership in the UK. Nevertheless, the

potential problem of sample selection that might have biased estimation results of previous

empirical studies which used the LCF survey data is clearly visible from our split of the

sample between home-owner occupiers and renters. Campbell and Cocco (2007) attempted

Fig. 3 The frequency and percentage distribution of household expenditure ratios in the UK in 2010

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

to deal with this sampling error by assuming that the decision to become a homeowner or

renter is endogenous, and hence correlated with individual characteristics such as income,

consumption, age and economic status. Indeed they found that over time for a fixed birth

year, the group of tenants in their dataset shrank and became more concentrated in the low-

income population. Deaton (2000) remarked that quantile regressions which we employ in

this study are appropriate for exploring such potential shifts in household survey data.

3.3 Housing Tenure and Expenditure Ratios Across the UK Regions

Figure 5 provides an overview of the aggregate weekly spending relative to the income of

our responding families subdivided into the aforementioned three types of housing occu-

pation across the four regions of the UK. Four interesting features emerge from this chart.

The first is the fact that an average property owner in all regions of the UK, with the

exception of Northern Ireland, is consuming more than his/her average weekly wage. This

is especially pronounced in England where a typical house owner-occupier spent roughly

40.90 % in excess of average weekly wage in 2010. We note from Fig. 1 that the

expenditure ratio for the cohort of English homeowners was more than double the fig-

ure observed for the UK as whole from 2000 to 2010. This abnormality was doubtless

influenced by the high level of house prices in the South East and London in particular

(Disney et al. 2002; Medland 2011). Campbell and Cocco (2007) uncovered a similar

heterogeneity in the consumption effect of house prices with the highest estimated coef-

ficient for old homeowners than for young homeowners and for renters. They remarked

that the observed differentiation in the magnitude of expenditure ratios was linked to the

fact that the vast majority of older homeowners were paying off fixed rate mortgages which

remained relatively high. Additionally, Crossley et al. (2012) commented on the resilience

Fig. 4 The distribution of respondents in the survey by housing tenure

J. C. Nwachukwu

123

Author's personal copy

of household spending on certain durable goods and services, including council tax, utility

bills and home improvement which were customarily paid by property owners and land-

lords. It seemed that the majority of families in England went into debt or drew down on

their financial assets in order to settle payment on these household items in 2010.

Second, the difference in the mean expenditure ratio of owner-occupiers and renters in

England was considerably greater than the corresponding variation in the other parts of the

UK. For example, the dissimilarity in the spending ratio for property owners in England

was 23 percentage points and 52 percentage points higher than that reported for private and

public tenants in that order. Campbell and Cocco (2007) remarked on the greater severity

of borrowing constraints faced by renters who tend to have lower assets than their

landlords.

Third, contrary to expectation, respondents who are renting from public authorities in

Scotland and Wales consumed a larger proportion of their incomes compared with private

tenants in the same regions. This may be a reflection of the greater overall social security

provided by councils and the devolved assemblies in these regions. It could be that Welsh

and Scottish public tenants judge it less necessary to save for old age and unemployment

compared to the, perhaps, more self-reliant and better educated private renters.

Fourth, in Northern Ireland, there was little difference in the expenditure behaviour of

our three different categories of house occupancy. This region is the poorest part of the UK

following decades of civil unrest. We may therefore propose that social housing pre-

dominates here, forcing down private rental and ownership costs in the other two sub-

groups. Interestingly, the expenditure ratios for all types of housing tenure in Northern

Ireland was at, or below 100 %, perhaps due to precautionary savings associated with the

uncertainty caused by decades of civil disorder in the region.

Fig. 5 The average for the weekly expenditure to gross income ratio in 2010

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

4 Regression Model and Expected Relationships

The purpose of this section is twofold. First, it specifies the regression model underlying

the empirical analysis. Second, it describes the expected impact on total expenditure ratios

of selected household characteristics.

4.1 Model Specification

The linear model which we use to capture the influence of the characteristics of households

on the expenditure ratio can be broadly expressed as follows.

LCYi ¼ a0 þ b0 Xið ÞDþp 0 REGIONð ÞDþx

0 Zið ÞDþei ð2Þ

The variable LCYi is a natural logarithm of our aggregate expenditure ratio for respondent i.

The decision to take the natural log of the series follows from the skewness in the distribution

of expenditure ratio in Fig. 3. Asteriou and Hall (2007), Brooks (2008) and Caglayn and Astar

(2012) suggested that the use of such logarithmic-linear functional forms may help resolve

misspecification errors, including those arising from non-normality and heteroscedasticity of

residuals. Moreover, the use of logarithmic models means that the estimated differential slope

coefficients may be interpreted as marginal propensities of consumption.

The term Xi is an N 9 1 matrix of dummy variables Dð Þ representing the covariate of primary interest—homeownership status of respondent i. REGION is UK region where the

respondent claims to live at the time of the LCF survey. The symbol Zi is an N 9 8 matrix

of conditioning variables drawn from a pool of potential household characteristics theo-

retically or empirically linked to changes in expenditure ratio in the economics literature

(e.g., Miles 1997; Campbell and Cocco 2007; Jacobson et al. 2010; van de Ven 2011;

Meng et al. 2014). The definitions of these explanatory variables are provided in Table 1.

For ease of interpretation, all the categories for our chosen household characteristics are

binary dummy variables. The symbols b0 p0 and x0 are the differential slope coefficients to be estimated. The term a0 is a constant which captures the expected value of the expen- diture ratio of the household category omitted from the regression because it was assigned

a value of zero in the construction of the dummy variables. The notation ei is an idiosyncratic residual capturing omitted determinants of LCYi, including measurement

error to which data from surveys are particularly prone. It is expected to have a zero mean

and constant variance.

4.2 Bivariate Analysis and Expected Relationships

To be able to initially describe the potential linkage between the expenditure ratio and each

of our selected binary dummy variables Xið ÞD; REGIONð ÞD and Zið ÞD � �

in isolation, we ran

a series of bivariate regressions based on the expression in Eq. 2. The null hypothesis to be

tested is that the average consumption ratio of those respondents in our sample with the

characteristics assigned a value of one (say home owner-occupiers) does not deviate sig-

nificantly from the mean ratio of their peers within the category with an allocated value of

zero, (say the renters). As a result, it is proposed that the value of the differential slope

coefficients b0; p0 and x0 is equal to 0 at the conventional 5 % confidence level. Columns 1–10 of Table 3 present the result for each of our ten household attributes. The interpre-

tation of these pairwise correlation coefficients is presented under (i) variables of interest

and (ii) control variables.

J. C. Nwachukwu

123

Author's personal copy

T a b le

3 O L S b iv a ri a te

re g re ss io n o f th e d e te rm

in a n ts o f U K

e x p e n d it u re

ra ti o

C o lu m n 1

C o lu m n 2

C o lu m n 3

C o lu m n 4

C o lu m n 5

C o lu m n 6

C o lu m n 7

C o lu m n 8

C o lu m n 9

C o lu m n 1 0

In te rc e p t

4 .4 0 * * *

[3 2 4 .8 8 ]

4 .3 3 * * *

[2 2 3 .8 3 ]

4 .4 9 * * *

[3 2 1 .0 7 ]

4 .3 2 * * *

[2 6 7 .6 1 ]

4 .4 2 * * *

[3 2 4 .5 6 ]

4 .4 0 * * *

[4 4 0 .0 4 ]

4 .4 0 * * *

[3 5 5 .9 8 ]

4 .3 5 * * *

[4 2 2 .6 9 ]

4 .4 0 * * * [2 8 4 .2 6 ]

4 .3 3 * * *

[4 4 1 .1 1 ]

T E N U R E

- 0 .0 8 4 * * *

[- 5 .0 4 4 ]

– …

… …

… …

… …

…

R E G IO

N …

0 .0 2 6

[1 .2 5 8 ]

– …

… …

… …

… …

W A G E S

… –

- 0 .2 5 0 * * *

[- 1 5 .4 2 6 ]

– …

… …

… …

…

IN T E R N E T

… –

… 0 .0 4 2 * * *

[2 .2 7 5 ]

– …

… …

… …

E M P L O Y M E N T

… –

– –

- 0 .1 2 3 * * *

[- 7 .3 9 6 ]

– …

… …

…

C L A S S

– –

– –

… - 0 .1 7 2 * * *

[- 1 1 .4 1 3 ]

… …

… …

G E N D E R

… –

– –

– –

- 0 .0 8 2 * * *

[- 5 .0 9 8 ]

… …

…

H S IZ E

… –

– –

– –

– - 0 .0 0 3

[- 0 .2 2 1 ]

… …

A D U L T S

… –

– –

– –

– –

- 0 .0 8 6 * * *

[- 4 .8 2 0 ]

…

C H IL D R E N

… –

– –

– –

– –

… 0 .0 5 9 * * *

[3 .6 7 5 ]

N o o f o b se rv a ti o n s

5 2 6 2

5 2 6 2

5 2 6 2

5 2 6 2

5 2 6 2

5 2 6 2

5 2 6 2

5 2 6 2

5 2 6 2

5 2 6 2

D e p e n d e n t v a ri a b le

is th e n a tu ra l lo g a ri th m

o f th e ra ti o o f a g g re g a te

e x p e n d it u re

to in c o m e fo r th e U K . N u m b e rs

in […

.] b ra c k e t a re

t st a ti st ic s. T h e sy m b o l * * * in d ic a te s

si g n ifi c a n c e a t th e 1 %

c o n fi d e n c e le v e l. T h e d e fi n it io n o f a ll v a ri a b le s a re

p ro v id e d in

T a b le

1

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

4.2.1 Variables of Interest

The estimated coefficients for our two variables of primary focus—house ownership and

regional location—are in Columns 1 and 2 respectively. The first variable (TENURE)

relates to the marginal consumption propensities of households with different ownership

status for their main residences. Following from the identity in Eq. 1, we may expect that

the consumption ratio should rise as the share of respondents that live in their own resi-

dences increase, if new borrowing against financial and housing collateral becomes easier

(Campbell and Cocco 2007). The statistically significant negative coefficient (-0.08)

indicates otherwise. What is more, the negative correlation contradicts the figures reported

in our descriptive statistics in Table 2 and Fig. 5. Such a revision in the sign on the

differential slope coefficients highlights the importance of the unobservable determinants

of the spending ratio captured by the common intercept term a0 in Eq. 2. Taken together, the negative coefficient on (TENURE) infers that individuals who lived in their own

properties had lower average expenditure ratios than renters in 2010. This finding is

consistent with the wealth effect of the lower house prices at the end of recession. Such

would have intensified the borrowing constraints on home owner-occupiers vis-à-vis

renters. It is possible that these owner-occupiers might have chosen to use their savings

and/or windfalls from the government to reduce outstanding debt stock in order to lower

their cost of mortgage debt servicing. In any event, homeowners have got to be such as a

consequence of an inherent desire to save for down and annual payments on mortgages

(Sheiner 1995).

The positive coefficient for the second series REGIONS in Column 2 is statistically

insignificant. The implication is that the total expenditure ratio of a typical family in any

part of the UK, say England, is comparable with their counterparts in Scotland, Wales and

Northern Ireland. This may be due to some extent to the universal dispensations of the

welfare state, particularly child and unemployment benefit, as well as the winter fuel

allowance for the elderly. Medland (2011) noted that the relatively high expenditures by

London residents were substantially offset by the lower spending in the other regions in

England, especially those living in the lowest expenditure counties in Yorkshire, the North

East and the Humber region.

4.2.2 Control Variables

The pairwise correlations between each of the eight conditioning dummy variables in the

matrix Zi and our total consumption ratios are reported in Columns 3–10. Two covariates

are predicted to show a significantly positive coefficient, indicating an increase in

household total expenditure ratio. These attributes are: (i) households with internet con-

nection (INTERNET) and (ii) the number of children in a family unit (CHILDREN). There

is a suggestion that a 10 percentage point increase in the share of families with internet

connections would raise the UK average spending ratio by 0.42 % rising to 0.59 % for the

fraction of households with children relative to those without. A possible explanation may

be related to the life-cycle theory in the sense that the majority of our internet users, as well

as those individuals with dependents under the age of sixteen, are likely to be young people

who are expected to have a higher marginal propensity to consume than the elderly. What

is more, a positive coefficient could be taken as a sign that internet connection in a

household is private rather than a shared or public good. In fact, most individuals in a

household have personal mobile internet contracts with different providers. Additionally,

J. C. Nwachukwu

123

Author's personal copy

the ease of comparing prices and shopping for items on line through mobile communi-

cation equipment, such as smartphones and tablet computers, could have increased the

amount of goods and services purchased by consumers of all ages, especially the young.

The remaining six household characteristics in our conditioning set have a negative

effect on the aggregate expenditure ratio. These are presented here in a decreasing order of

the absolute size of the estimated differential slope coefficients.

The first is the source of earnings for the household reference person (WAGE) with an

estimated coefficient of (-0.25) in Column 3. The inverse relationship between the

spending ratio and the proportion of households who derive their income largely from a

regular wage or salary is consistent with the rationale that this category of earners has

higher incomes than pensioners and those on benefits. Thus, it is expected that they will

have a lower marginal propensity to consume. Then too, it is possible that the uncertainty

surrounding pay conditions and hours worked at the end of the financial crisis increased

considerably compared with state and index linked pensions, in particular. Risk-averse

families fearing prospective unemployment or a cut in their wages might wish to maintain

some buffer by raising the amount of their precautionary savings from current income.

Indeed, descriptive statistics in Table 2 indicate that this discrepancy in the spending ratio

was largest among the residential homeowner category for which borrowing constraints

were exaggerated at the end of the crisis period.

The second important conditioning variable is a dummy that captures the effect on total

expenditure ratio of an increase in the percentage of families headed by respondents who

claim to work in managerial positions (CLASS). As expected, we estimated a negative

correlation coefficient of (-0.172) in Column 6 in line with the lower consumption

propensities for this higher income class. Alternatively, from the asset pricing model, we

may infer that the lower mean expenditure ratio for our managerial class is related to the

fact that they are better educated and so are more likely to hold financial assets such as

shares and bonds. Therefore, the fall in asset prices at the end crisis period would have

reduced their perceived existing wealth. Such could have inhibited their relative

expenditures.

The third prominent attribute relates to the effect of employment status (EMPLOY-

MENT) on the spending ratio. The predicted negative coefficient (-0.12) in Column 5 may

be associated with a greater scepticism on the part of our working families about their

future job security and pay increases. We may suppose that in the aftermath of financial

crisis, many British workers expected their overall disposable incomes to fall significantly

below their then current pay packets in the foreseeable future due to the higher rate of

unemployment. Such would have motivated them to cut consumption expenditures in line

with the predictions of the permanent income theory. Another reason may be that those in

employment, especially in full time work, are unlikely to be free at the hours when shops

are open and the time of year when the prices of recreational activities including holidays

abroad are discounted. Such could have led to a decline in their aggregate expenditure

ratio.

The fourth relevant conditioning series is the size of households in terms of the number

of adults (ADULTS) in Column 9. A negative coefficient of (-0.086) posits that the

average expenditure ratio of families with more than two adults is lower than that for their

equivalents with a single or two persons aged 16 and over. This outcome is presumably

because all these adults are likely to be earning or receiving welfare benefits, leading to a

high overall household income. Then too, cost savings arising from discounts are enjoyed

by larger adult families that buy items such as food, holidays and insurance in bulk (Deaton

and Paxson 1998). Moreover, gains allied to opportunities to share goods as well as to

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

spread fixed family payments, such as those for mortgage or rent, telephone rentals,

utilities, council tax and a car, should be higher for larger adult households (Jacobson et al.

2010).

The significantly negative coefficient (-0.082) on the fifth important conditioning

dummy variable (GENDER) in Column 7 indicates a substantial disparity in the mean

expenditure ratio of households headed by males versus females. The underlying reason for

the lower spending ratio observed for our male respondents is that men are more likely to

be sole earners and in full employment with higher incomes than females. They would

have handed over a larger percentage of their income to their wives for general household

expenditure. Consequently, we may conjecture that the majority of males who responded

to the LCF questionnaire failed to accurately record their contributions to family budgets in

their expenditure diaries, while the recipient females did so. Then too, males may be more

likely than females to shop on line, set up direct debit payments and have credit and/or

debit cards because of their employment and income positions (Pahl 1999, 2000). Pay-

ments of utility and shopping bills with such cashless methods often attract considerable

discounts, leading to a lower aggregate spending bill for men.

The overall size of a household (HSIZE) is measured in terms of the number of people,

including children, living together in a family unit. The insignificantly negative correlation

coefficient of (-0.003) in Column 8 suggests that the mean for the total expenditure ratio

for a family with more than three persons is comparable to the ratio for smaller sized

households. This finding could be taken as a signal that individuals in larger homes are

increasingly substituting private with shared goods (Jacobson et al. 2010). For example, it

is reasonable to suppose that at the end of the recession, families, especially those with

more than three people, gradually replaced meals in restaurants, pubs and take-aways with

food cooked at home. Also, cars, clothing and children’s toys are more likely to be shared

among family members in the wake of a crisis, leading to a reduction in household

expenditure bills relative to income. Nonetheless, the fact that the estimated coefficient is

insignificantly different from zero at the conventional five-percent level implies that we do

not have sufficient information in our aggregate expenditure data to determine what goods

and services were cut back and to what extent by our large versus small households.

To summarise, results of the pairwise correlation analysis support our decision to dif-

ferentiate between the spending habits of home owner-occupiers versus renters. However,

we find little heterogeneity in the response of household consumption ratio to regional

fixed effects. The analysis in the subsequent section attempts to further explore the sen-

sitivity of these findings to estimation method and a simultaneous addition of the other

household characteristics in our conditioning set.

5 Multivariate Regression Method and Main Results

The bivariate regressions in the previous section provide a simplified account of the extent

to which expenditure-to-income ratio for an average UK family in 2010 is predicted by

each of our chosen eleven characteristics in isolation. However, there are instances where

interactions might exist between our set of household attributes. For example, represen-

tations in Fig. 5 suggest that individuals who claimed to live in their own properties in

England have considerably higher expenditure ratios compared with other home owners

and renters in the other UK regions. A multivariate regression is therefore required to

establish the correlation between the average expenditure ratio and such interrelatedness

J. C. Nwachukwu

123

Author's personal copy

between housing tenure and regional location of respondents over and above the effects of

all our other household characteristics enumerated in the vectors Xi and Zi. Specifically, the

extended regression model which we analyse in this paper may be represented as follows:

LCYi ¼ a0 þ b1ðTENUREiÞD þ p1ðREGIONÞD þ c1ðTENUREi � REGIONÞD þ x1ðWAGEiÞD þ x2ðINTERNETiÞD þ x3ðEMPLOYMENTiÞD þ x4ðCLASSiÞD þ x5ðGENDERiÞD þ x6ðHSIZEiÞD þ x7ðADULTSiÞD þ x8 CHILDRENið ÞDþei

ð3Þ

To estimate Eq. 3, we utilise a quantile regression approach which minimises the sum

of absolute error. Such median estimators are increasingly used in the econometric

literature in place of conditional mean models, such as OLS, as a convenient way for

providing a more complete description of the underlying distribution of a response

variable.

Quantile regression was introduced by Koenker and Bassett (1978) in an attempt to

extend the classical least squares ideas to the estimation of conditional quantiles of a

dependent variable given a set of control regressors. The model expresses the conditional

distribution of a response variable into quantile or percentile of the observed covariates.

For the present study, we split our sample of 5263 households into 20 percentiles of equal

size according to their observable characteristics which are captured by each of the

regressors in Eq. 3. Koenker and Hallock (2001) remarked that the use of such a relatively

large number of distinct cells is more efficient than non-parametric approaches which are

traditionally employed in tests for the distributional robustness of conditional mean

models.

Detailed representations of the conditional quantile function, which are usually

minimized by researchers in order to obtain the vector of parameters bs; are provided in the articles by Koenker and Bassett (1978), Buchinsky (1998), Deaton (2000),

Koenker and Hallock (2001) and Ronning and Schulze (2004). However, for ease of

computation and interpretation, a simplified form of a conditional quantile regression of

a random variable Y given K independent regressors is employed. This may be written

as follows:

Quantileh YijKið Þ ¼ chKi þ ehi ð4Þ

where Yi is the natural logarithm of the total expenditure to income ratio for respondenti, K

is a vector of the covariates listed in Eq. 3. The term ch is a vector of parameter coefficients and h is the quantile being analysed. We estimated the conditional expenditure ratio for nineteen separate quantiles {0.05…0.95} using the least absolute deviation (LAD) esti- mator in EVIEWs Version 8. The LAD estimator maintains the classical linear regression

assumption that the error terms are independent and identically distributed (i.i.d). Standard

errors were obtained using the bootstrapping option recommended by Buchinsky (1998).

The results are presented in Table 4.

To clarify our discussion here, Fig. 6 presents a summary of quantile regression results

for all of our chosen covariates bounded within a 95 % confidence interval. For each of the

regressors, we plot the nineteen distinct quantile regression estimates for tau (s) ranging from 0.05 to 0.95. The point estimate may be interpreted as the impact on the total

expenditure ratio of a 1 percentage point change of the covariate, holding the other

regressors fixed. So each plot indicates the quantile or s scale on the horizontal axis and the marginal effect of the covariate on the vertical axis.

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

T a b le

4 Q u a n ti le

re g re ss io n e st im

a te s

V a ri a b le s

In te rc e p t

T E N U R E

R E G IO

N S

T E N R E G N S

W A G E

IN T E R N E T

E M P L Y M N T

C L A S S

G E N D E R

H S IZ E

A D U L T S

C H IL D R E N

0 .0 5

3 .5 4 2 * * *

[6 2 .4 4 ]

- 0 .0 8 5

[- 1 .2 7 1 ]

- 0 .0 6 5

[- 1 .0 9 4 ]

0 .0 3 8

[0 .4 7 6 ]

- 0 .2 3 3 * * *

[- 4 .5 7 1 ]

0 .1 8 5 * * *

[5 .4 9 2 ]

0 .1 4 0 * * *

[2 .7 5 9 ]

- 0 .0 3 8

[- 1 .1 4 3 ]

- 0 .0 3 9

[- 1 .3 2 2 ]

- 0 .0 9 5 * *

[- 2 .1 9 4 ]

0 .0 9 3 * * *

[2 .8 0 3 ]

0 .1 6 0 * * *

[3 .9 2 8 ]

0 .1

3 .7 1 8 * * *

[7 1 .4 2 ]

- 0 .1 0 1 *

[- 1 .6 6 1 ]

- 0 .0 1 0

[- 0 .1 7 6 ]

0 .0 2 1

[0 .2 9 8 ]

- 0 .2 3 6 * * *

[- 4 .9 7 2 ]

0 .2 1 0 * * *

[7 .0 8 6 ]

0 .0 9 1 * *

[1 .9 7 1 ]

- 0 .0 4 7

[- 1 .6 2 5 ]

- 0 .0 4 2 *

[- 1 .8 7 9 ]

- 0 .0 1 5

[- 0 .2 7 7 ]

0 .0 6 8 * *

[2 .3 9 5 ]

0 .0 9 3 * *

[1 .9 2 6 ]

0 .1 5

3 .8 5 8 * * *

[9 0 .0 7 ]

- 0 .0 9 4 * *

[- 2 .1 1 7 ]

0 .0 0 5

[0 .1 2 8 ]

0 .0 1 2

[0 .2 5 7 ]

- 0 .2 6 3 * * *

[- 7 .0 0 2 ]

0 .1 7 8 * * *

[6 .5 6 8 ]

0 .1 0 4 * * *

[2 .9 5 6 ]

- 0 .0 5 4 * *

[- 2 .5 9 2 ]

- 0 .0 4 8 * *

[- 2 .1 9 4 ]

0 .0 1 7

[0 .5 2 1 ]

0 .0 5 0 * *

[1 .9 7 1 ]

0 .0 7 3 * *

[2 .1 8 8 ]

0 .2

3 .9 5 6 * * *

[9 5 .3 1 ]

- 0 .0 9 0 * *

[- 1 .9 7 7 ]

0 .0 1 2

[0 .2 8 8 ]

0 .0 1 7

[0 .3 8 5 ]

- 0 .2 5 1 * * *

[- 7 .5 1 6 ]

0 .1 6 3 * * *

[6 .3 5 3 ]

0 .0 8 2 * *

[2 .5 5 7 ]

- 0 .0 5 8 * * *

[- 2 .7 2 9 ]

- 0 .0 5 0 * *

[- 2 .5 2 4 ]

0 .0 0 2

[0 .0 5 9 ]

0 .0 3 3

[1 .5 7 2 ]

0 .1 0 1 * * *

[3 .7 3 0 ]

0 .2 5

4 .0 7 8 * * *

[1 0 0 .3 5 ]

- 0 .1 3 0 * * *

[- 2 .6 4 7 ]

- 0 .0 1 6

[- 0 .3 9 2 ]

0 .0 5 7

[1 .2 1 1 ]

- 0 .2 7 3 * * *

[- 7 .4 5 8 ]

0 .1 5 1 * * *

[5 .4 5 3 ]

0 .0 7 6 * *

[2 .3 0 2 ]

- 0 .0 4 0 * * *

[- 1 .8 8 8 ]

- 0 .0 5 0 * *

[- 2 .4 5 6 ]

0 .0 1 5

[0 .5 7 5 ]

0 .0 4 0

[1 .5 8 2 ]

0 .0 8 0 * * *

[3 .2 3 8 ]

0 .3

4 .1 3 8 * * *

[1 0 4 .3 2 ]

- 0 .0 8 8 * *

[- 1 .9 9 4 ]

0 .0 0 4

[0 .1 1 2 ]

0 .0 2 3

[0 .5 3 6 ]

- 0 .2 7 9 * * *

[- 8 .6 2 9 ]

0 .1 1 7 * * *

[4 .6 3 7 ]

0 .0 9 7 * * *

[3 .1 4 0 ]

- 0 .0 4 3 * *

[- 2 .4 3 7 ]

- 0 .0 4 9 * * *

[- 2 .9 0 9 ]

0 .0 0 8

[0 .3 1 9 ]

0 .0 3 1

[1 .5 2 7 ]

0 .0 9 8 * * *

[3 .7 8 5 ]

0 .3 5

4 .2 2 8 * * *

[1 2 1 .1 9 ]

- 0 .1 2 0 * * *

[- 3 .1 8 2 ]

- 0 .0 2 5

[- 0 .7 2 7 ]

0 .0 6 6 *

[1 .8 1 7 ]

- 0 .3 0 3 * * *

[- 1 0 .9 7 2 ]

0 .1 2 2 * * *

[5 .8 3 5 ]

0 .1 0 5 * * *

[3 .9 9 7 ]

- 0 .0 4 0 * * *

[- 2 .8 9 8 ]

- 0 .0 4 9 * * *

[- 3 .2 7 1 ]

0 .0 1 2

[0 .4 3 1 ]

0 .0 2 2

[1 .1 6 7 ]

0 .0 9 2 * * *

[3 .4 2 9 ]

0 .4

4 .3 0 2 * * *

[1 2 4 .0 3 ]

- 0 .1 3 3 * * *

[- 3 .4 4 3 ]

- 0 .0 3 6

[- 1 .0 6 7 ]

0 .0 8 4 * *

[2 .2 4 9 ]

- 0 .3 1 5 * * *

[- 1 0 .1 6 8 ]

0 .1 1 6 * * *

[5 .5 9 0 ]

0 .1 0 3 * * *

[3 .3 3 6 ]

- 0 .0 4 1 * *

[- 2 .5 9 5 ]

- 0 .0 4 7 * * *

[- 3 .2 1 7 ]

0 .0 2 9

[1 .0 6 8 ]

0 .0 1 1

[0 .6 1 4 ]

0 .0 8 8 * * *

[3 .2 4 4 ]

0 .4 5

4 .3 3 7 * * *

[1 6 0 .8 3 ]

- 0 .1 0 0 * * *

[- 2 .6 3 8 ]

- 0 .0 2 3

[- 0 .8 2 7 ]

0 .0 5 9

[1 .5 5 3 ]

- 0 .3 3 3 * * *

[- 1 0 .5 7 0 ]

0 .1 1 9 * * *

[5 .0 1 1 ]

0 .1 1 0 * * *

[3 .2 4 3 ]

- 0 .0 3 3 * *

[- 2 .0 3 8 ]

- 0 .0 4 6 * * *

[- 2 .9 9 7 ]

0 .0 3 0

[1 .0 8 2 ]

0 .0 0 4

[0 .1 8 6 ]

0 .0 9 0 * * *

[3 .2 8 9 ]

0 .5

4 .3 7 6 * * *

[1 7 3 .7 9 ]

- 0 .0 7 5 * * *

[- 2 .3 1 2 ]

- 0 .0 0 7

[- 0 .2 4 9 ]

0 .0 3 5

[1 .0 4 9 ]

- 0 .3 5 5 * * *

[- 1 1 .3 0 1 ]

0 .1 2 8 * * *

[5 .3 4 5 ]

0 .1 1 8 * * *

[3 .5 2 2 ]

- 0 .0 2 4

[- 1 .4 4 9 ]

- 0 .0 4 8 * * *

[- 3 .3 8 2 ]

0 .0 2 2

[0 .9 1 8 ]

0 .0 1 3

[0 .5 7 9 ]

0 .0 8 1 * * *

[3 .5 0 0 ]

0 .5 5

4 .4 1 9 * * *

[1 5 4 .6 6 ]

- 0 .0 6 1 *

[- 1 .6 6 7 ]

0 .0 0 7

[0 .2 2 6 ]

0 .0 2 3

[0 .5 9 9 ]

- 0 .3 5 2 * * *

[- 1 3 .7 7 5 ]

0 .1 4 6 * * *

[6 .4 2 7 ]

0 .1 0 6 * * *

[3 .5 9 2 ]

- 0 .0 2 7

[- 1 .5 6 0 ]

- 0 .0 3 5 * *

[- 1 .9 4 9 ]

0 .0 2 6

[1 .1 2 5 ]

- 0 .0 0 8

[- 0 .3 4 0 ]

0 .0 7 5 * * *

[3 .1 4 2 ]

0 .6

4 .4 7 6 * * *

[1 1 6 .0 4 ]

- 0 .0 4 1

[- 0 .9 4 4 ]

0 .0 2 7

[0 .6 7 6 ]

0 .0 0 1

[0 .0 2 6 ]

- 0 .3 4 5 * * *

[- 1 1 .1 7 0 ]

0 .1 2 7 * * *

[4 .9 9 3 ]

0 .1 1 0 * * *

[3 .1 0 6 ]

- 0 .0 5 0 * * *

[- 2 .6 9 3 ]

- 0 .0 5 0 * *

[- 2 .7 6 1 ]

0 .0 3 5

[1 .0 9 5 ]

- 0 .0 0 3

[- 0 .1 1 4 ]

0 .0 6 5 * *

[1 .9 8 1 ]

0 .6 5

4 .5 5 0 * * *

[1 0 2 .9 5 ]

- 0 .0 5 0

[- 1 .1 6 8 ]

0 .0 1 5

[0 .3 8 0 ]

0 .0 1 4

[0 .3 0 9 ]

- 0 .3 5 4 * * *

[- 9 .0 4 0 ]

0 .1 3 4 * * *

[5 .3 5 0 ]

0 .1 2 2 * * *

[2 .8 2 7 ]

- 0 .0 7 7 * * *

[- 4 .1 1 3 ]

- 0 .0 5 4 * *

[- 2 .9 0 0 ]

0 .0 3 8

[1 .1 9 6 ]

- 0 .0 1 3

[- 0 .5 3 4 ]

0 .0 7 3 * *

[2 .0 8 3 ]

0 .7

4 .6 2 8 * * *

[1 2 2 .7 9 ]

- 0 .0 6 3 *

[- 1 .7 1 1 ]

0 .0 0 7

[0 .1 9 7 ]

0 .0 2 8

[0 .7 2 7 ]

- 0 .3 5 0 * * *

[- 9 .5 4 2 ]

0 .1 4 4 * * *

[6 .4 2 4 ]

0 .1 1 2 * * *

[2 .8 9 4 ]

- 0 .0 8 3 * * *

[- 4 .7 6 4 ]

- 0 .0 6 2 * * *

[- 3 .4 9 4 ]

0 .0 4 2

[1 .4 9 4 ]

- 0 .0 2 5

[- 1 .0 1 5 ]

0 .0 7 8 * *

[2 .5 6 0 ]

0 .7 5

4 .7 1 9 * * *

[1 1 8 .4 8 ]

- 0 .0 6 3

[- 1 .3 9 6 ]

0 .0 0 0

[0 .0 1 3 ]

0 .0 4 3

[0 .9 5 3 ]

- 0 .3 7 3 * * *

[- 8 .6 6 9 ]

0 .1 6 0 * * *

[6 .0 1 9 ]

0 .1 1 0 * *

[2 .5 8 7 ]

- 0 .0 7 3 * * *

[- 3 .8 9 1 ]

- 0 .0 6 2 * * *

[- 3 .4 2 2 ]

0 .0 2 9

[0 .9 7 0 ]

- 0 .0 5 1 * *

[- 1 .9 7 3 ]

0 .0 7 6 * *

[2 .5 1 6 ]

J. C. Nwachukwu

123

Author's personal copy

T a b le

4 c o n ti n u e d

V a ri a b le s

In te rc e p t

T E N U R E

R E G IO

N S

T E N R E G N S

W A G E

IN T E R N E T

E M P L Y M N T

C L A S S

G E N D E R

H S IZ E

A D U L T S

C H IL D R E N

0 .8

4 .7 7 7 * * *

[1 0 3 .6 8 ]

- 0 .0 0 8

[- 0 .1 4 5 ]

0 .0 4 9

[1 .2 4 9 ]

- 0 .0 1 6

[- 0 .3 1 4 ]

- 0 .4 0 5 * * *

[- 7 .5 4 6 ]

0 .1 6 9 * * *

[6 .0 7 2 ]

0 .1 3 7 * * *

[2 .7 8 6 ]

- 0 .0 9 9 * * *

[- 4 .7 5 7 ]

- 0 .0 2 4

[- 1 .0 5 9 ]

0 .0 0 5

[0 .1 5 6 ]

- 0 .0 8 7 * * *

[- 2 .6 7 4 ]

0 .0 7 6 * *

[2 .8 9 6 ]

0 .8 5

4 .9 4 0 * * *

[8 4 .6 2 ]

- 0 .0 3 0

[- 0 .4 9 9 ]

0 .0 1 5

[0 .2 5 7 ]

0 .0 1 5

[0 .2 4 5 ]

- 0 .4 4 8 * * *

[- 9 .4 6 9 ]

0 .1 7 8 * * *

[5 .1 9 6 ]

0 .1 4 9 * * *

[3 .3 7 1 ]

- 0 .1 3 6 * * *

[- 5 .2 0 6 ]

- 0 .0 2 6

[- 1 .3 7 4 ]

- 0 .0 1 0

[- 0 .2 8 2 ]

- 0 .1 0 0 * * *

[- 3 .0 5 6 ]

0 .0 5 2 *

[1 .7 7 7 ]

0 .9

5 .1 2 1 * * *

[6 0 .1 0 ]

- 0 .0 2 2

[- 0 .2 5 4 ]

0 .0 1 7

[0 .2 1 0 ]

0 .0 0 2

[0 .0 2 9 ]

- 0 .4 9 1 * * *

[- 8 .3 6 2 ]

0 .1 7 6 * * *

[3 .3 6 0 ]

0 .1 9 2 * * *

[3 .4 1 9 ]

- 0 .2 2 1 * * *

[- 7 .1 7 9 ]

- 0 .0 3 4

[- 1 .2 4 2 ]

0 .0 0 0

[0 .0 0 0 ]

- 0 .1 4 4 * * *

[- 3 .5 6 6 ]

0 .0 1 9

[0 .7 5 5 ]

0 .9 5

5 .4 7 2 * * *

[5 1 .6 1 ]

- 0 .0 2 7

[- 0 .2 0 9 ]

0 .0 0 1

[0 .0 0 7 ]

0 .0 1 0

[0 .0 8 0 ]

- 0 .5 7 5 * * *

[- 5 .3 5 2 ]

0 .2 0 7 * * *

[3 .4 9 8 ]

0 .2 8 1 * * *

[2 .8 3 7 ]

- 0 .3 3 9 * * *

[- 5 .8 3 6 ]

- 0 .0 9 9 *

[- 1 .9 1 4 ]

- 0 .1 0 2 *

[- 1 .8 2 9 ]

- 0 .1 7 5 * * *

[- 2 .7 5 4 ]

0 .0 9 5

[1 .6 3 2 ]

T h e e st im

a te d re g re ss io n m o d e l is sp e c ifi e d in

E q . 3 u si n g th e le a st a b so lu te d e v ia ti o n (L A D ) e st im

a to r in

E V IE W s V e rs io n 8 ; D e p e n d e n t v a ri a b le L C Y i is th e n a tu ra l lo g a ri th m

o f th e ra ti o o f a g g re g a te

e x p e n d it u re

to in c o m e fo r th e U K ; T h e in d e p e n d e n t v a ri a b le s c o m p ri se

a ll th e b iv a ri a te

d u m m ie s d e fi n e d in

T a b le

1 ; N u m b e rs

in […

.] b ra c k e t a re

t st a ti st ic s. T h e sy m b o ls * * * , * * a n d * in d ic a te si g n ifi c a n c e a t th e 1 , 5 a n d 1 0 %

c o n fi d e n c e le v e l re sp e c ti v e ly . T h e e st im

a te d c o e ffi c ie n ts a re

d if fe re n c e s in

th e m a rg in a l ra te o f

c o n su m p ti o n b e tw e e n h o u se h o ld s in

th e c a te g o ry

a ss ig n e d th e v a lu e o f o n e a n d th o se

w it h th e a tt ri b u te

a ll o c a te d a v a lu e o f z e ro

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

The signs on the coefficients on the house tenure dummy (TENURE) are insensitive to

the concurrent inclusion of our chosen conditioning set in the sense that they retain their

negative values for all quantiles. This finding provides support for the effectiveness of the

incentives provided by the Bank of England and the UK government in an attempt to

promote demand for both new and existing houses across all sections of society.

The parameter coefficient for the regional dummy (REGION) is persistently insignifi-

cant at all quantiles, even though the English households at the upper quantile appear to

have a slightly higher spending ratio than their counterparts in the other UK regions.

The coefficient on our interaction variable (TENREGNS) which was included to capture

any potential disparity in the consumption habits of homeowners in England and their

peers in the rest of the UK has the expected positive sign, although the estimated slope

differential of (?0.084) at the 40th percentile is the only statistically significant figure. This

perhaps captures the fact that English middle classes aspire to higher priced housing, cars,

private education and more expensive foreign holidays. Also, Disney et al. (2002) sug-

gested that the size of the estimated coefficients on these regional dummy variables cor-

relates closely and positively with average regional house prices.

With respect to our selected conditioning variables, we found that the sign on the

coefficients is consistent with those originally reported using the pairwise correlation

analysis in Sect. 3.2.2. The only important exceptions where we obtained a revision in the

sign for the estimated coefficient were for the following two categories: (i) employment

status of household head and (ii) the number of adults in a household.

The coefficient on the EMPLOYMENT variable reverted to a significantly positive sign

across all quantiles. The effect is at its strongest at the upper quantile, rising from 0.1 for

the lower percentile to 0.3 for the 95th quantile. The inference is that these later households

have lower incomes and so have higher marginal consumption rates. They are therefore

3.0

3.5

4.0

4.5

5.0

5.5

6.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

C

-.3

-.2

-.1

.0

.1

.2

.3

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

TENURE

-.3

-.2

-.1

.0

.1

.2

.3

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

REGIONS

-.3

-.2

-.1

.0

.1

.2

.3

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

TENREGNS

-.8

-.6

-.4

-.2

.0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

WAGE

.05

.10

.15

.20

.25

.30

.35

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

INTERNET

.0

.1

.2

.3

.4

.5

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

EMPLYMNT

-.5

-.4

-.3

-.2

-.1

.0

.1

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

CLASS

-.25

-.20

-.15

-.10

-.05

.00

.05

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

GENDER

-.3

-.2

-.1

.0

.1

.2

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

HSIZE

-.4

-.3

-.2

-.1

.0

.1

.2

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

ADULTS

-.05

.00

.05

.10

.15

.20

.25

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Quantile

CHILDREN

Fig. 6 Quantile process estimates

J. C. Nwachukwu

123

Author's personal copy

expected to spend increasing proportions of their additional earnings from employment on

private goods and services rather than substitute them for shared or public goods.

The results also show that the expenditure ratio of households consisting of at least three

adults at the upper 80th percentile is considerably lower than those at the bottom and

middle quantiles. As we said earlier, the underlying reason for this is the economies of

scale arising from gains to be derived from shared goods among adults in a family unit,

particularly those on lower incomes.

The intercept of the model is significantly positive for all quantiles. Such an outcome

may be taken as evidence of a higher total expenditure ratio for those categories which

were assigned a value of zero and so were excluded from the regression in order to avoid

the problem of a dummy variable trap.

To evaluate the quality of a quantile regression model, EViews produces a series of

goodness of fit measures. They include: (i) an adjusted R-squared which is analogous to

that reported from conventional OLS regression analysis. We obtained an adjusted

R-square which indicates that almost 42 % of the variation in the ratio of aggregate

expenditure to income was explained by our choice of independent variables. (ii) the

statistics for an equality test which compares the slope coefficient for the median against

the estimated upper and lower quantiles. We observed that a Chi square statistic of 83.25 is

statistically significant. The implication is that the estimated slope coefficients differ

considerably across conditional quantile values and (iii) the statistic for a test for the

degree of symmetry for the parameter coefficients around the median quantile. An esti-

mated Chi square statistic of 16.16 with associated p value of (0.1839) is taken as evidence

that the null of conditional symmetrical quantiles around the median cannot be rejected.

6 Conclusions and Policy Recommendations

This paper is primarily concerned with the spending behaviour of those individuals who

live in their own properties in the UK at the end of the recent recession in 2010. The study

uses data from the LCF 2010 survey of 5263 respondents who consistently kept a daily

record of their household income and spending on specified categories of goods and

services at regular 2 week intervals in that year. The key findings with related policy

actions which were uncovered from our bivariate and multivariate quantile regression

models are as follows:

First, the aggregate spending-to-income ratio of home owner-occupiers across the UK is

significantly lower than for renters, particularly for households in the middle quantile. This

finding gives support to the Bank of England’s decision to keep the basic interest rate at

0.5 % in order to improve credit conditions. Also, the latest government ‘‘Help to Buy

Equity Loans and Mortgage Guarantees’’ scheme introduced in March 2013 should alle-

viate the credit constraint on would-be home owners and so encourage them to increase

their effective demand for housing, furniture and home appliances with a consequent

increase in aggregate output and employment.

Second, we find insignificant variation in the aggregate spending of families which live

in the different UK regions but which are similar in other respects. This lack of regional

discrepancy may be a sign that targeted welfare benefits to support low income families

with children and disabilities, together with the elderly and unemployed has succeeded in

narrowing the gap in income and expenditure on essential items such as food, clothing,

housing and heating across the country. For example, the elderly especially those aged 70

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

and above, throughout the UK are entitled to a free bus pass, television licences and

pharmaceutical prescriptions. They also receive state pensions, winter fuel allowances

linked to inflation, as well as payments for care homes where they have insufficient

personal assets.

Third, access to internet connections and employment status are the two characteristics

in our conditioning set which were found to be most important in raising household

expenditure ratios, judging by the absolute size, statistical significance and persistence of

their estimated parameter coefficients across all quantiles.

With respect to internet connection, we many infer that government policies to dereg-

ulate the broadband market for the provision of superfast internet services to homes should

encourage more competition and cut the cost of shopping online. Then too, efforts to combat

cybercrime and to strengthen the legal protection afforded to online shoppers should

enhance general confidence, especially among the retired population. Moreover, the gov-

ernment should support schemes to provide free access to Wi-Fi in towns and in libraries

coupled with computer training for the unemployed in particular. Another initiative to

promote sustainable online shopping includes publicly-funded advertising campaigns on

television, bill boards and newspapers to publicize the availability of free price comparison

websites, as well as the organisations which help people to switch providers of items such as

utilities and mortgages which absorb a significant proportion of family budget.

In terms of general employment, it is recommended that the authorities pay particular

attention to actions which enhance expectations for long-term well-paid jobs by lower-

income families which rely mainly on a regular wage or salary. An example of such

policies is the stance adopted by the government to maintain fiscal discipline. This now

appears to be raising the overall economic growth with employment, as well as pay and

consumer spending. Besides the recent increase in the minimum wage for adults aged 25

and above from £6.31 at present towards the so-called living wage of £8.80 for London and

£7.65 for the rest of the country is encouraging. Supporters of the living wage campaign

have argued that the government should name and shame firms which do not pay wages

that enable their employees to live above the poverty line.

An important weakness of this study is that the data on expenditure is for all goods and

services and for all types of homeowners. An examination of the individual components of

these aggregates for mortgage, negative equity and non-mortgage owner-occupiers would

provide a better explanation of what goods and services were cut back to the greatest extent

by each type of property owner and the reasons lying behind their spending decisions.

Another limitation of this study is the fact that the empirical results are obtained from a

quantile regression function. However, there is a growing debate in the literature that non-

parametric methods, such as Neural Network which do not require researchers to impose a

priori a functional form on the estimates, are more apposite for dealing with both outliers

of the dependent variables and the lack of information about the correct functional form. A

key area for further research therefore would be to re-estimate our extended regression

models using such non-parametric techniques. The results could then be compared with our

benchmark least absolute deviation estimator.

Acknowledgments I am grateful for help with the data here to Augustine Nwachukwu and most particu- larly to a reviewer from this journal for their detailed comments and suggestions.

J. C. Nwachukwu

123

Author's personal copy

References

Abel, A. B. (1985). Precautionary saving and accidental bequests. American Economic Review, 75, 777–791.

Abel, A. B. (1986). The failure of Ricardian equivalence under progressive wealth taxation. NBER working paper no. 1983. Cambridge, MA: National Bureau of Economic Research.

Aitchison, J., & Brown, J. A. C. (1954). A synthesis of Engel curve theory. The Review of Economic Studies, 22(1), 35–46.

Anderson, M., Bechhofer, F., & Gershuny, J. (1994). The social and political economy of the household. Oxford: Oxford University Press.

Asteriou, D., & Hall, S. G. (2007). Applied econometrics: A modern approach using Eviews and Microfit (Revised ed.). London: Palgrave Macmillan.

Atkinson, A. B., Gomulka, J., & Stern, N. H. (1990). Spending on alcohol: Evidence from the family expenditure survey 1970–1983. Economic Journal, 100(402), 808–827.

Attanasio, O., & Weber, G. (1994). The UK consumption boom of the late 1980’s: Aggregate implications of microeconomic evidence. Economic Journal, 104(427), 1269–1302.

Attanasio, O. P., & Weber, G. (2010). Consumption and saving: Models of intertemporal allocation and their implications for public policy. Journal of Economic Literature, 48, 693–751.

Barro, R. (1974). Are government bonds net wealth? Journal of Political Economy, 81, 1095–1117. Belsky, E., & Prakken, J. (2004). Housing wealth effects: Housing’s impact on wealth accumulation, wealth

distribution and consumer spending. National Centre for real estate research report W04-13, Harvard University, Boston.

Benito, A., Thompson, J., Waldron, M., & Wood, R. (2006). House prices and consumer spending. Bank of England Quarterly Bulletin, Summer, 142–154.

Bernheim, D. B. (1987). Ricardian equivalence: An evaluation of theory and evidence. NBER Macroeco- nomics Annual, 2, 263–316. http://www.nber.org/chapters/c11103

Bernheim, B., & Bagwell, K. (1986). Is everything neutral? Journal of Political Economics, 96(2), 308–338. Berry, S., & Williams, R. (2009). Household savings, bank of England. Quarterly Bulletin Household Sector

Article. http://www.bankofengland.co.uk/publications/Documents/quarterlybulletin/qb090302.pdf. Blanchard, O. J. (1985). Debt, deficits and finite horizons. Journal of Political Economy, 93(2), 223–247. Blundell, R. (1988). Consumer behaviour: Theory and empirical evidence—A survey. The Economic

Journal, 98(389), 16–65. Blundell, R., Pashardes, P., & Weber, G. (1993). What do we learn about consumer demand patterns from

micro data? American Economic Review, 83(3), 570–597. Bostic, R., Gabriel, S., & Painter, G. (2009). Housing Wealth, Financial Wealth, and Consumption: New

Evidence from Micro data. Regional Science and Urban Economics, 39, 79–89. Bosworth, B., Burtless, G., & Sabelhaus, J. (1991). The decline in saving: Some microeconomic evidence.

Brookings Paper on Economic Activity, 26, 313–336. Brooks, C. (2008). Introductory econometrics for finance (2nd ed.). Cambridge: Cambridge University

Press. Brumberg, R. E. (1956). An approximation to the aggregate saving function. Economic Journal, 46, 66–72. Buchinsky, M. (1998). Recent advances in quantile regression—A practical guide for empirical research.

Journal of Human Resources, 33(1), 88–126. Caglayn, E., & Astar, M. (2012). A microeconometric analysis of household consumption expenditure

determinants for rural and urban areas in Turkey. American International Journal of Contemporary Research, 2(2), 27–34.

Campbell, J. (1987). Does saving anticipate declining labour income? An alternative test of the permanent income hypothesis. Econometrica, 55(6), 1249–1273.

Campbell, Y., & Cocco, F. (2007). How do house prices affect consumption? Evidence from micro data. Journal of Monetary Economics, 54, 591–621.

Capps, O., & Love, L. (1983). Determinants of household expenditure on fresh vegetables. Southern Journal of Agricultural Economics, 15, 127–132.

Carmichael, J. (1982). On Barro’s theorem of debt neutrality: The irrelevance of net wealth. American Economic Review, 72(1), 202–213.

Carroll, C. (2001). A theory of the consumption function, with and without liquidity constraints. Journal of Economic Perspectives, 15(3), 23–45.

Carroll, C. (2004). Housing wealth and consumption expenditure. Paper prepared for the Academic Con- sultants meeting of the Board of Governors of the Federal Reserve System.

Carruth, A., & Henley, A. (1990). The housing market and consumer spending. Fiscal Studies, 11(3), 27–38.

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

Case, K. E., Quigley, J. M., & Shiller, R. J. (2005). Comparing wealth effects: The stock market versus the housing market. Advances in Macroeconomics, 5(1), 1534–6013.

Chen, J. (2006). Re-evaluating the association between housing wealth and aggregate consumption: New evidence from Sweden. Journal of Housing Economics, 15, 321–348.

Collis, J., Grayson, A., & Joha, S. (2010). Econometric analysis of alcohol consumption in the UK, HMRC working paper 10.

Crossley, T. F., Low, H., & O’Dea, C. (2012). Household consumption through recent recessions. Institute of Fiscal Studies, working paper 11/18.

Crossley, T. F., & O’Dea, C. (2010). The wealth and saving of UK families on the eve of the crisis. Institute for Fiscal Studies report no. 71.

Danziger, S., Van Der Gaag, J., Smolensky, E., & Taussig, M. (1982). The life-cycle hypothesis and the consumption behavior of the elderly. Journal of Post Keynesian Economics, 5(2), 208–227.

Deaton, A. (1986). Life-cycle models of consumption: Is the evidence consistent with the theory? NBER working paper no. 1910.

Deaton, A. (2000). The analysis of household surveys: A microeconomic approach to development policy. Baltimore, MD: The John Hopkins University Press.

Deaton, A., & Paxson, C. (1998). Economies of scale, household size, and the demand for food. Journal of Political Economy, 106, 897–930.

Diamond, P. A. (1965). National debt and neoclassical economic growth. American Economic Review, 55, 1125–1150.

Disney, R., Henley, A., & Jevons, D. (2002). House price shocks, negative equity and household con- sumption in the UK in the 1990s. Wales: Manuscript, School of Management and Business, University of Wales Aberystwyth.

Duesenberry, J. S. (1949). Income, saving and the theory of consumer behaviour. Cambridge: Harvard University Press.

Dvornak, N., & Kohler, M. (2003). Housing wealth, stock market wealth and consumption: A panel analysis for Australia. Reserve Bank of Australia research discussion paper, RDP2003-07.

Dynan, K., & Maki, D., (2001). Does the Stock Market Wealth Matter for Consumption? Working paper, Division of Research and Statistics, Board of Governors of the Federal Reserve System, Washington, DC, 2001–2023.

Edison, H., & Slok, T. (2001). Wealth effects and the new economy. Working paper, international monetary fund.

Elliot, J. W. (1980). Wealth and wealth proxies in a permanent income model. Quarterly Journal of Economics, 95, 509–535.

Engel, J., & Kneip, A. (1996). Recent approaches to estimating Engel curves. Journal of Economics, 63(2), 187–212.

Engelhardt, G. V. (1996). House prices and home owner saving behaviour. Regional Science and Urban Economics, 26(3/4), 313–336.

Feldstein, M. (1986). The effects of fiscal policies when incomes are uncertain: A contradiction to Ricardian equivalence. Working paper no. 2062, NBER, Cambridge, MA.

Friedman, M. (1957). A theory of the consumption function. Princeton: Princeton University Press. Gale, W. G., & Sabelhaus, J. (1999). The savings crisis: In the eye of the beholder?. Washington, DC: The

Milken Institute. Hall, R. E. (1978). Stochastic implication of the life cycle-permanent income hypothesis: Theory and

evidence. Journal of Political Economy, 86, 971–988. Hall, R. E. (1981). Intertemporal substitution in consumption. National Bureau of Economic Research,

Working paper no. 720. NBER Cambridge, MA. Härdle, W., Hildenbrand, W., & Jerison, M. (1991). Empirical evidence on the law of demand. Econo-

metrica, 59, 1525–1549. Hirsch, F. (1976). Social limits to growth. New York: The Twentieth Century Fund. Houthakker, H. S. (1952). The econometrics of family budgets. Journal of the Royal Statistical Society,

Series A, 115(1), 1–28. Houthakker, H. S. (1957). An international comparison of household expenditure patterns, commemorating

the century of Engel’s law. Econometrica, 25, 532–551. Jacobson, D., Mavrikiou, P. M., & Minas, C. (2010). Household size, income and expenditure on food: The

case of Cyprus. The Journal of Socio-Economics, 39, 319–328. Khoman, E., & Weale, M. (2008). Are we living beyond our means? A comparison of France, Italy, Spain

and the United Kingdom. National Institute discussion paper no. 311. Koenker, R., & Bassett, G. (1978). Regression quantiles. Econometrica, 46(1), 33–50.

J. C. Nwachukwu

123

Author's personal copy

Koenker, R., & Hallock, K. F. (2001). Quantile regression. Journal of Economic Perspectives, 15(4), 143–156.

Kotlikoff, L. J. (1988). Intergenerational transfers and savings. Journal of Economic Perspectives, 2(2), 41–58.

Kotlikoff, L., & Spivak, A. (1981). The family as an incomplete annuities market. Journal of Political Economy, 89(2), 372–391.

Leland, H. (1968). Saving and uncertainty: The precautionary demand for saving. The Quarterly Journal of Economics, 82(3), 465–473.

Lettau, M., & Ludvigson, S. (2004). Understanding trend and cycle in asset values: Re-evaluating the wealth effect on consumption. American Economic Review, 94, 276–299.

Lise, J., & Seitz, S. (2011). Consumption inequality and intra-household allocations. Review of Economic Studies, 78(1), 328–355.

Mankiw, N. G. (1981). The permanent income hypothesis and the real interest rate. Economics Letters, 7, 307–311.

Mankiw, N. G., & Shapiro, N. D. (1985). Trends, random walks and tests of the permanent income hypothesis. Journal of Monetary Economics, 16, 165–174.

Manning, W. G., Blumberg, L., & Moulton, L. H. (1995). The demand for alcohol: The differential response to price. Journal of Health Economics, 14, 123–148.

Mayer, T. (1972). Permanent income, wealth and consumption. Berkeley: University of California Press. Medland, A. (2011). Regional Family Spending in England. Office for National Statistics. http://

neighbourhood.statistics.gov.uk/HTMLDocs/images/Family%20Spending%20Regional_tcm97-109491. pdf.

Meng, Y., Brennana, A., Purshouseb, R., Hill-McManusa, D., Angusa, C., Holmesa, J., et al. (2014). Estimation of own and cross price elasticities of alcohol demand in the UK—A pseudo-panel approach using the living costs and food survey 2001–2009. Journal of Health Economics, 34(2014), 96–103.

Miles, D. (1992). Housing markets, consumption and financial liberalisation in the major economies. European Economic Review, 36, 1093–1136.

Miles, D. (1993). House prices, personal sector wealth and consumption: some conceptual and empirical issues. The Manchester School, 61, 35–59.

Miles, D. (1997). A household level survey of the determinants of incomes and consumption. The Economic Journal, 107(440), 1–25.

Millard, S., & Power, J. (2004). The effects of stock market movements on consumption and investment: Does the shock matter? Bank of England working paper no. 236.

Miller, D. (1998). A theory of shopping. London: Polity Press. Modigliani, F. (1961). Long-run implications of alternative fiscal policies and the burden of the national

debt. Economic Journal, 71, 730–755. Modigliani, F. (1963). The life-cycle hypothesis of savings: Aggregate implications tests. American Eco-

nomic Review, 53, 53–84. Modigliani, F. (1964). The life-cycle hypothesis of savings: A correction. American Economic Review,

54(2), 111–113. Modigliani, F. (1966). The life cycle hypothesis of saving, the demand for wealth and the supply of capital.

Social Research, 33(2), 160–217. Modigliani, F., & Ando, A. (1957). Tests of the life-cycle hypothesis of saving. Bulletin of Oxford

University Institute of Statistics, 19, 99–124. Modigliani, F., & Sterling, A. (1976). Government debt, government spending and private sector behaviour:

A comment. American Economic Review, 76(5), 1168–1179. Muellbauer, J., & Murphy, A. (1997). Booms and busts in the UK housing market. The Economic Journal,

107(445), 1701–1727. Nationwide. (2011). House price index. http://www.nationwide.co.uk/hpi/. Nickell, S. (2004). Household debt, house prices and consumption growth. Bank of England Quarterly

Bulletin, Autumn, 383–389. O’Driscoll, G. P, Jr. (1977). The Ricardian non-equivalence theorem. Journal of Political Economy, 85,

207–210. Pahl, J. (1999). Invisible money: Family finances in the electronic economy. Bristol: Policy Press. Pahl, J. (2000). The gendering of spending within households. Radical Statistics, 75, 38–48. Poterba, J. M. (2000). Stock market wealth and consumption. The Journal of Economic Perspectives, 14(2),

99–118. Prais, S. J. (1952). Non-linear estimates of the Engel curves. The Review of Economic Studies, 20(2),

87–104.

Tenure and Spending Within UK Households at the End of the…

123

Author's personal copy

Purshouse, R. C., Meier, P. S., Brennan, A., Taylor, K. B., & Rafia, R. (2010). ‘‘Estimated effect of alcohol pricing policies on health and health economic outcomes in England: An epidemiological model. Lancet, 375, 1355–1364.

Ronning, G., & Schulze, N. (2004). A microeconomic characterisation of household consumption using quantile regression. Applied Economics Quarterly, 50(2), 183–208.

Scholz, J. K., Seshadri, A., & Khitatrakun, S. (2006). Are Americans saving ‘‘optimally’’ for retirement? Journal of Political Economy, 114, 607–643.

Scitovsky, T. (1976). The Joyless economy: An inquiry into human satisfaction and con-sumer dissatis- faction. Oxford: Oxford University Press.

Seater, J. J. (1993). Ricardian equivalence. Journal of Economic Literature, 31, 142–190. Sheiner, L. (1995). Housing prices and the savings of renters. Journal of Urban Economics, 38, 94–125. Sinai, T., & Souleles, N. (2005). Owner-occupied housing as a hedge against rent risk. Quarterly Journal of

Economics, 120, 763–789. Skinner, J. (1989). Housing wealth and aggregate saving. Regional Science and Urban Economics, 19,

305–324. Stigler, G. J. (1954). The early history of empirical studies of consumer behaviour. Journal of Political

Economy, 62(2), 95–113. Tanner, J. E. (1978). Fiscal policy and consumer behaviour. Review of Economics and Statistics, 61,

317–321. Thaler, R. (1990). Anomalies: Saving, fungibility, and mental accounts. The Journal of Economic Per-

spectives, 4(1), 193–205. Van de Ven, J. (2011). Expenditure and disposable income trends of UK households: Evidence from micro-

data. National Institute Economic Review, No. 218.

J. C. Nwachukwu

123

Author's personal copy