Planning for Housing in the Post-Barker Era: Affordability,
Household Formation and Tenure Choice
By
Geoffrey Meen* and Mark Andrew**
*International Centre for Housing and Urban Economics, Department of Economics,
The University of Reading Business School,
PO Box 219, Whiteknights, Reading RG6 6AW, England.
Telephone: 0118 378 6029
Fax: 0118 378 6533
Email: [email protected]
** Faculty of Finance
Cass Business School
City University
London EC1Y 8TZ
ICHUE Paper No. 8
August 2007
The work described in this paper was conducted for the Department of Communities and Local
Government. The model is also used by the National Housing and Planning Advice Unit. However, the
views expressed here are those of the authors and not necessarily those of the Government or NHPAU.
1. Introduction
For many years, official projections of household formation have provided the basis for
regional and local housing plans. In turn, official projections are based on expected
population distributions by age, marital status and gender and detailed estimates of
headship rates or household representative rates, HRR, for each group. However, the
HRRs are trend projections, based on census and Labour Force Survey data, and do not
take into account structural changes in economic or institutional conditions that may
change past trends. In particular, the projections do not attempt to capture the effects of
improving or worsening affordability on the ability of individuals to form new
households. The most recent official household projections look ahead to 2029. The
estimates are based on 2004 population figures and indicate an annual average increase
between the two years of approximately 220,000 households per annum for England as a
whole, with the increase weighted towards the southern regions. Since England currently
constructs approximately 160,000 new dwellings per annum and the government, until
recently, was committed to raising the figure to 200,000 dwellings by 2016, there was
prima facie a case that a crude housing shortage would occur. The 2007 Green Paper,
which raised the housing target to 240,000 units was a recognition of the likely shortage.
In fact, a simple national number count is over-simplistic. First, it takes no account of
spatial patterns or current shortages/surpluses. Second, it takes no account of the market.
In some sense, the household projections are an estimate of future need not market
demand. Potential shortages can be eliminated by price adjustment, although there may,
of course, be undesirable social externalities, for example overcrowding, leading to wider
social pathologies. It is because of the externalities that housing is considered a merit
good and successive governments have made a commitment to decent housing, although
the form of the commitment has varied over time. Third, a number count takes no
account of housing quality; all units are considered equivalent, although four-bedroom
detached houses in the suburbs yield a different level of housing services from an urban
two-room flat.
There are no official estimates of the expected distribution of households by tenure,
although this is clearly important for the planning of social housing provision.
1
Furthermore, private rental investors in the so-called Buy-to-Let market need estimates of
likely future demand. The government also has an aspiration to raise the owneroccupation rate; consequently, it needs estimates of the expected outcome under existing
economic conditions and constraints. But, whereas the provision of a decent home,
independently of a household’s economic circumstances, may be justified as a merit
good, the same issue does not hold with tenure; the tenure distribution is largely a market
outcome and, arguably, governments should attempt to maintain tenure neutrality in
terms of intervention, unless there are associated externalities. It has been suggested, for
example, that home-ownership has positive externalities by raising neighbourhood
quality, since neighbourhood quality is capitalised into house prices, e.g. DiPasquale and
Glaeser (1999), Rohe et al (2002). Alternatively, ownership may have negative macro
externalities if it reduces labour mobility.
Amongst the many recommendations of the Barker Review of Housing Supply (Barker,
2004) was the call for the adoption of affordability targets at the national and regional
level. This recognised that the planning system is currently unresponsive to market price
signals and argued that more land needed to be brought forward if affordability worsened
beyond some, undefined, target level. Although Barker did not recommend a single
affordability indicator, in accepting the recommendations, the Government is
concentrating on the ratio of lower quartile house prices to earnings.
What has not been widely recognised, however, is that affordability targets have
profound implications for traditional approaches to modelling household formation and
housing need. It is no longer possible to rely on extrapolations of past trends for HRRs.
The aim of affordability targets is to change those past trends. Similarly, affordability
targets imply that matching housing construction levels to the number of households is no
longer adequate. In this paper, we demonstrate, that typically this will not stabilise
affordability. Generally, net additions to the housing stock, which include conversions
and changes in property uses, as well as new construction, would have to be higher than
the number of new households in order to account for the higher demand for housing
services by existing households as their incomes rise over time. The Barker proposals
2
have been attacked as implying that large amounts of Greenfield land in the South would
need to be covered. Furthermore it has been suggested that the proposals would
contribute to the decline of city centres. In fact, as we attempt to demonstrate, neither of
these criticisms is valid since neither is necessarily implied by affordability targets.
However, regional affordability targets do require the greater use of formal econometric
models. In particular, estimates are needed of:
(i)
(ii)
(iii)
The responsiveness of house prices (and, therefore, affordability) to increases
in housing construction. This, in turn, requires estimates of the income and
price elasticities of housing demand.
The responsiveness of household formation to improvements in affordability.
The responsiveness of regional migration flows to relative affordability
differentials.
Meen (2005) concentrated on (i) and, here, the emphasis is on (ii). The likely tenure
distribution of households requires a further range of elasticities. More precisely:
(i)
(ii)
(iii)
The responsiveness of tenure to relative cost differentials
The impact of credit market constraints
The effects of supply shortages
Once again, however, affordability targets imply that simple extrapolations of past trends
are no longer adequate.
Section 2 considers the traditional approach to projecting household numbers and
compares the estimates to the most recent regional plans for net additions to the housing
stock. Section 3 develops an alternative formal model of household formation and tenure
choice, more appropriate to the post-Barker framework. As noted above, past trends
cannot be simply extrapolated since the aim of the targets is to change the trends. In
Section 4, the model addresses the issue of the expected outcome for household
formation under market outcomes. But this does not necessarily imply that household
formation will meet the official estimates. Section 5 uses the model to discuss a number
of current policy questions, for example the possibility of raising home-ownership rates
and home-ownership sustainability. We also discuss the implications of increasing
3
housing supply in the light of the Barker Review. This includes the commonly-expressed
view that England does not need large numbers of new homes because of existing
vacancies and the contentious issue of demolitions. This leads on to a brief consideration
of city regeneration versus Greenfield new building. Section 6 draws conclusions.
2. The Traditional Approach
The most recent English national and regional projections of household formation use, as
a starting point, estimates of population in 2004. The projections look ahead to 2029;
national figures are summarised in Table 1 and the regional disaggregations in Table 2. A
number of features stand out. First, between 2004 and 2029, the total number of
households is expected to increase by 5.4 million or approximately 220,000 per annum.
By comparison, the earlier 2003-based household projections indicated an expected
annual increase of approximately 210,000 per annum between 2003 and 2026. Even
earlier 1996-based projections showed an increase of 293,000 per annum between 1996
and 2021 (the equivalent figure for 1996-2021 on the latest projections is 400,000). For
2021, the estimates of total households are now almost 1 million higher than in the 1996based projections. Therefore, there has been an upward drift in the official projections
over time. Second, the outturn for household formation over the last ten years was
considerably over-estimated by the 1996-based household projections. This overprediction coincided with a worsening of housing affordability. Third, Table 1 shows a
continuing reduction in the average household size from 2.34 persons in 2004 to 2.09 in
2029. Fourth, 3.8 million of the increase between 2004 and 2029 are amongst single
person households, with an absolute reduction in the number of married couples. Fifth,
although not shown in Table 1, of the 155,000 annual increase in single person
households between 2004 and 2026, only 14,400 per annum are expected to be
households with heads under the age of 35. Sixth, as Table 2 shows, the increases are
concentrated in the South at least in absolute terms. Power and Houghton (2007, page
139), using 1996-based household figures show shortages of new building relative to
household formation in the southern regions, but excess supplies of construction in the
Midlands and, particularly, the northern regions. This pattern is unlikely to have changed
4
on the basis of the latest household projections. These six points enter into the later
discussion.
Table 1. 2004-Based Household Projections
England
Household types:
married couple
cohabiting couple
lone parent
other multi-person
one person
2004
2016
2021
2026
2029
9,527,000
1,987,000
1,591,000
1,421,000
6,536,000
9,050,000
2,944,000
1,830,000
1,629,000
8,384,000
8,978,000
3,204,000
1,882,000
1,708,000
9,200,000
8,898,000
3,424,000
1,928,000
1,775,000
9,951,000
8,832,000
3,552,000
1,949,000
1,817,000
10,347,000
All households
21,062,000
23,837,000
24,973,000
25,975,000
26,497,000
Private household population
Average household size
49,200,000
2.34
52,331,000
2.20
53,625,000
2.15
54,787,000
2.11
55,381,000
2.09
Table 2. 2004-Based Regional Household Projections
All households
England
North East
North West
Yorks & Humber
E Midlands
W Midlands
East
London
South East
South West
2004
21,062,000
1,095,000
2,895,000
2,122,000
1,799,000
2,206,000
2,304,000
3,112,000
3,368,000
2,160,000
2029 Change 2004-29
% change
26,497,000
5,435,000
25.80
1,275,000
180,000
16.44
3,507,000
612,000
21.14
2,703,000
581,000
27.38
2,296,000
497,000
27.63
2,657,000
451,000
20.44
2,954,000
650,000
28.21
4,078,000
966,000
31.04
4,211,000
843,000
25.03
2,817,000
657,000
30.42
5
Table 3. Vintages of Official Household Projections (000s)
England
1996
1997
1998
1999
2000
2001
2002
2003
2004
2006
2011
2016
2021
2026
2029
2004 based
19,727
19,816
19,924
20,052
20,222
20,523
20,720
20,904
21,062
2003 based**
19,727
19,816
19,924
20,052
20,222
20523
20714
20904
21098
1996 based*
20186
20371
20555
20740
20865
20990
21138
21286
21434
21,519
22,646
23,837
24,973
25,975
26,497
21485
22566
23705
24781
25713
21730
22520
23310
24000
2004 based-1996 based
-459
-555
-631
-688
-643
-467
-418
-382
-372
-211
126
527
973
* 1997, 1998, 2000, 2002-4 are interpolations
** 2002, 2004 are interpolations
As noted above, the HRRs underlying Tables 1 and 2 are projected on the basis of past
trends. Therefore, implicitly, they assume that the factors determining HRRs also
continue their previous trends. For example, the literature indicates that the probability
that any individual will form a separate household is related both to demographic
characteristics and to housing affordability (Andrew and Meen 2003). Consequently,
official projections assume that affordability continues over the future at a similar
position to the past. But the aim of the Barker affordability targets is to change and
improve past trends. Other things equal, this would raise HRRs relative to past trends.
Furthermore, Section 3 suggests that, should the Barker proposals be unsuccessful in
raising the supply of housing services, affordability could worsen relative to the past.
Under either scenario, an extrapolation of past HRRs is not appropriate. We have already
noted that an extrapolation of trends from a 1996 base led to an over-prediction of
households. The paper demonstrates later that the official estimates are a special case
under particular assumptions concerning affordability.
Table 4 turns to housing supply and considers housing targets under Regional Planning
Guidance and the proposed revisions under the draft Regional Spatial Strategies. These
figures are based on the expected increase in the number of households. But the plans
6
take years to reach fruition and, consequently, the RPG plans were based on the 1996based household projections published in 1999. We have already seen that household
estimates have been raised in recent years, adding to potential housing shortages. The
RSS draft revisions attempt to take this into account, but, given the lags in setting
regional plans, construction is always chasing a moving target. An important caveat to
this is the Government announcement in the 2007 Housing Green Paper of a new target
of 240,000 new homes per annum by 2016, which are not reflected in the figures in Table
4. But, even taking this expected increase into account, prima facie there is a case for a
system that is more immediately responsive to market pressures.
Table 4 is based on net additions to the housing stock, not gross construction of new
homes. In fact as Table 5 shows, net additions occur through a combination of new
building, conversions, changes in property use and demolitions. Therefore, increases in
the housing stock do not have to use Greenfield sites out of existing towns, but can also
occur through regeneration of the existing stock within urban areas. In itself, calls for
large increases in net additions to the stock, do not imply “concreting over the South
East”, since all forms of net additions add to the stock of housing services, although in
different ways. In fact, Table 5 shows that, for England as a whole, demolitions are
approximately equal to the sum of “conversions” and “change of use”, so that gross
completions are similar in number to net additions. But the values for London show that
there is no reason why this should necessarily hold, since net additions considerably
exceed new construction activity. Again the moral is that net additions to the housing
stock should not necessarily be equated with new completions.
.
Table 4. Plans for Annual Net Additions to the Housing Stock (Nos)
Finalised RPG Draft RSS
Plans
Plans
5,729
6,580
North East
12,790
22,844
North West
14,765
19,266
Yorks & Humber
13,700
20,418
E Midlands
15,156
n.a.
W Midlands
20,850
25,399
East
19,048
27,596
London
28,050
28,900
South East
20,200
22,978
South West
7
Table 5. The Composition of Net Additions (England and London, Numbers)
2000/01 2001/02
2002/03
2003/04 2004/05 2005/06
England
Housebuilding completions
Conversions (net)
Change of use (net)
Non-permanent dwelling additions (net)
Demolitions
133518
2108
9343
89
20050
129992
3040
15297
6
26256
137977 143958 155893
4009
5053
5624
17829 16442
15631
490
623
0
23200 20155 20091.5
163398
7296
19048
0
20424
Net additions
125008
122079
137105 145921
169318
London
2000/01 2001/02
2002/03
157057
2003/04 2004/05 2005/06
Housebuilding completions
Conversions (net)
Change of use (net)
Non-permanent dwelling additions (net)
Demolitions
14755
488
1582
15
2429
14053
586
2656
52
2012
15908
892
1845
149
1606
19394
1733
2691
-212
1243
24063
1437
2437
0
1816
18809
2254
2792
0
2894
Net additions
14411
15335
17188
22363
26121
20961
Nationally, equation (1) holds at all points in time. Any increase in new households must
be matched by a combination of an increase in new units, taking account of second
homes, changes in vacancies and demolitions. As already noted new units can occur
through conversions and changes in use as well as new building.
∆HH = ∆HS − ∆SEC − ∆VAC − DEM
HH
HS
SEC
VAC
DEM
=
=
=
=
=
(1)
number of households
number of housing units
second homes
vacancies
demolitions
Equation (1) is sometimes used to justify matching construction targets to the expected
rate of household formation, since second homes, vacancies and demolitions have
historically been small and stable in proportion to the housing stock. But in a market
system subject to affordability targets, the argument is inadequate. First, an important
reason why vacancies and demolitions have been low has been housing shortages and
consequent poor affordability. If housing costs are high, the opportunity cost of holding
properties vacant is also high. Similarly, if shortages keep house prices artificially high,
the average duration of property lives increases. But, under affordability targets, where
8
higher rates of new construction are likely, we would expect to observe accompanying
higher levels of vacancies and demolitions as part of the market adjustment. Second,
although (1) has to hold in terms of units, it takes no account of housing quality. As
demonstrated in the next section, even if HS=HH affordability will typically worsen,
because existing households will demand higher quantities of housing services as their
incomes rise. If the stock of services is inelastic, the real price of houses rises in
equilibrium over time. The point is that in a world of affordability targets, matching
numbers of households to numbers of units is no longer sufficient. In the next section, a
model is developed for housing planning more appropriate to the attainment of
affordability objectives. This is a system that is more responsive to market pressures than
the current plan-based approach, but recognises the role of housing as a merit good and
government objectives to provide access to decent housing.
3. An Alternative Approach
The section begins with a model, which defines a set of outcomes that might occur if
housing decisions were determined entirely by market processes. A simplified version is
portrayed in Figure 1. In the figure, the econometric relationships to be modelled are
given on the left-hand side – these cover inter-regional migration, the probability that any
individual will form a separate household, tenure choice and the demand for housing
services. The remainder of the figure defines the central aggregate outcomes – the total
number of households, the allocation of total households across the tenures, house prices,
rents and affordability, and vacancies/demolitions/second homes. Importantly, all the
variables on the left-hand side could be affected by changes in affordability and,
therefore, there is a link back to these variables from the aggregate level of affordability
(dotted line), where affordability is determined by market clearing processes.
Furthermore, if the future trend in affordability changes relative to the past, all the
variables in the model are affected.
The two boxes at the top and bottom of the figure define sets of variables that are
determined outside the structure of the model. These include the population at the end of
9
the previous year (t-1), births, deaths, international migration and average earnings. The
latter is the denominator in the government’s chosen definition of affordability (the ratio
of average house prices to earnings). There is evidence that earnings are affected by
housing market variables (e.g. Bover et al 1989, Blackaby and Manning 1992), but this
extension is ignored here. Similarly births, deaths and international migration may all be
affected by housing affordability, but we judge the likely elasticities to be fairly low and
the central ideas can be illustrated without the addition.
The bottom left-hand corner gives the supply of owner-occupier services. Although it
might be expected that the variable would be responsive to changes in house prices,
across the English regions, Meen (2006) indicates that the price elasticity of new housing
supply has fallen close to zero since the nineties. In the model, the supply of services is
treated as a policy instrument. In the Barker Review context, if affordability worsens
beyond a given target, further construction is brought on to the market. Note that “new
housing” is stated in italics above. This emphasises the fact that services can be increased
not only by new building, but also by conversions and renovations of the existing stock.
The equations in the left-hand column employ a combination of micro and aggregate
data. The probabilities of household formation and tenure choice are estimated on
individual data from the British Household Panel Survey, whereas the inter-regional
migration and owner-occupier housing service demand equations1 use regional timeseries information. In each of the central equations affordability is an important
determinant, but not the only influence. Therefore, Table 6 sets out the main2 influences
for each. The equations are discussed briefly in the following sub-sections, but it is
helpful to begin with the demand for housing services since it highlights a common
fallacy that is critical for achieving affordability targets.
1
As shown below, strictly regional house price equations are estimated from which the demand functions
can be inferred.
2
In fact, some of the equations include further influences, for example, the house price equations (see
Meen et al (2005) for full equation details). But those appearing in the table are the key factors for
determining the model’s properties.
10
Population (t-1)
+ (Births-Deaths)
+ International
Migration
Interregional
Migration
Population
of type (i)
Prob (individual of
type (i) forms
household type (j) )
Households
of type (j)
Prob (household of type
(j) is in each of the 3
tenures)
Demand for
housing services
by owners
Number of owning households
Number of private renters
Number of social renters
House
prices
Rents
Supply of owneroccupier housing
services
Vacancies,
demolitions,
second homes
AFFORDABILITY
Earnings
Figure 1. Flow Chart of the Model
11
Table 6. The Key Equations: Summary of the Main Influences
Influences
Demand for Housing Services (or house Number of households, the stock of
dwellings, real earnings, interest rates.
prices)
Marital status, age, gender, children, real
Probability of Household Formation
housing costs, real incomes, previous
household status.
Tenure costs, real incomes, credit
Tenure
conditions, previous tenure, marital status,
age, children, gender.
Relative house prices, housing availability,
Inter-regional Migration
relative earnings, unemployment.
3.1 The Demand for Owner-Occupier Housing Services
As discussed in the last section, the traditional planning approach attempts to relate
housing construction to the expected future number of households. But it is not the case
that this will necessarily stabilise affordability. The fallacy lies in treating all units as
identical, whereas different houses, in fact, contain different levels of housing services.
Consider the following housing services demand function (2), which is representative of
the literature.
ln( H d ) = a 0 + a1 ln(Y ) − a 2 ln( PH ) + a3 ln( HH ) − a 4 ln(r ) + ε 1
where:
Hd
Y
PH
HH
r
ε1
ln
ai
=
=
=
=
=
=
=
=
(2)
Aggregate demand for owner-occupier housing services
Real average earnings
Real house prices
Total number of households
Mortgage interest rate
error term
natural logarithm
set of coefficients (elasticities)
For a fixed (short run) supply of housing services, market equilibrium implies the house
price equation (3). A large literature estimates national price equations of similar form
(for example, Muellbauer and Murphy 1997, Meen and Andrew 1998a). A growing
literature also estimates similar equations at the regional level. This literature was
summarised in Meen and Andrew 1998b, although Cameron et al (2006) estimate a more
recent and more complex regional model, allowing for inter-regional price spill-overs.
12
ln( PH ) = (a 0 / a 2 ) + (a1 / a 2 ) ln(Y ) + (a3 / a 2 ) ln( HH ) − (a 4 / a 2 ) ln(r ) − (1 / a 2 ) ln( H s ) + ε 2
(3)
where
= Supply of housing services
Hs
If a3 = 1 (i.e. the demand for housing services rises proportionately to the number of
households), then equation (3) can be simplified to (4).
ln( PH ) = (a 0 / a 2 ) + (a1 / a 2 ) ln(Y ) − (a 4 / a 2 ) ln(r ) − (1 / a 2 ) ln( H S / HH ) + ε 2
(4)
In (4), a proportionate rise in both the housing stock and the number of households leaves
real house prices unchanged. This homogeneity property is supported by the empirical
studies noted above. But if homogeneity holds, then affordability, ln(PH/Y), is constant
only if (a1/a2) = 1, for given values of the other variables. If (a1/a2) > 1, then affordability
worsens over time (assuming growing incomes), even if HS=HH. However, (a1/a2) > 1
implies that the income elasticity of housing demand is higher than the price elasticity.
Both the above time-series studies find this to be the case. In our model, the income
elasticity is approximately one and the price elasticity -0.5, so (a1/a2) = 2. Therefore,
affordability worsens over time unless the supply of housing services rises faster than the
number of households or some other market stabiliser operates, for example, changes in
interest rates. But, of course, house prices are only one of the factors taken into account
by the Bank of England in setting interest rates.
In terms of the underlying economics, the equation implies that, as incomes rise, existing
households demand a higher quantity of housing services than they currently hold. This
might imply bigger houses in better neighbourhoods, for example. It may also involve a
tenure change. The model, therefore, introduces a form of filtering of the housing stock,
where higher income households move to higher quality homes and the homes they
vacate filter down to those further down the income distribution.
Equation (4) determines house prices in the model. But to determine the distribution of
households between tenures, assumptions are needed concerning relative housing costs in
each tenure. Because of past constraints on private rents and the administered nature of
13
social rent setting, it is not appropriate to estimate the relationships from historical data,
since most of the private rented sector is now uncontrolled. In a pure market system, it is
more relevant to assume a direct relationship between the three tenure prices. In a longterm setting, failure to do so will imply that all households will choose the cheapest
sector, for a property of a given quality, in the absence of other constraints (e.g. supply
shortages).
3.2 The Probability of Forming Separate Households
Given the distribution of population by age, gender, marital status, number of children
and income, the expected total number of households of each type can be obtained, using
the probability that individuals with each of the above characteristics will form a separate
household. Formally, these can be estimated from a probit model, using data from the
British Household Panel Survey. Unsurprisingly, those with the highest incomes are more
likely to form separate households, but in line with the literature (see Bramley et al
1997), the responsiveness of household formation to economic variables – both incomes
and prices – is fairly weak. The demographic influences are quantitatively much larger.
By contrast, in the tenure equations discussed below, economic factors are more
important than demographics. As a generalisation, demographics determine household
formation, but economics determines the tenure distribution.
Whether an individual currently lives in an independent household depends on her status
in the previous year. If she is already in a separate household, there is a high probability
that she will remain as a separate household. Alternatively, there is a low probability that
she will return to the parental home. By contrast, individuals who were previously living
with parents have a relatively low probability of forming a new household. In other
words, households exhibit a high degree of persistence and changes in income or housing
costs, for example, typically do not immediately lead to a change in status. However,
persistence leads to differential responses to changes in affordability across the regions.
The largest differences are in London. Because London has a young age distribution, a
smaller percentage of individuals have already formed separate households.
Consequently a worsening of affordability means that they remain with parents for
14
longer. The elasticity of household formation with respect to affordability is relatively
high. By contrast, regions that have older age distributions will have a higher percentage
of residents already in separate households. Given persistence, a worsening of
affordability is unlikely to persuade many households to give up their independent status.
Therefore, the elasticity of household formation with respect to affordability is relatively
weak.
Table 7 shows how household formation probabilities vary across individuals in London.
The second half of the table shows that, whatever the demographic status, if the
individual was previously in a separate household, she is likely to maintain that status.
Income (and housing costs, although not shown in the table) has little influence.
Alternatively shocks would have to be large to induce a change in status, for example
through divorce. But the first half of the table shows that the probabilities of household
formation are much lower for individuals, who were not previously independent. The
response to economic change (rows 4 and 5) is much larger. Consequently, regions that
have high percentages of the young, who have not already formed separate households,
will be more affected by variations in economic conditions.
Table 7 Probabilities of Household Formation (London, 2001)
Previously not separate household
Female, 25-29, single, no children,
quartile 4
Male, 20-24, single, no children,
quartile 2
Male, 30-34, single, no children,
quartile 4
Male, 30-34, partner, children,
quartile 4
Male, 30-34,
partner, children,
quartile 1
Previously separate household
Female, 25-29, single, no children,
quartile 4
Male, 30-34, single, no children,
quartile 4
Male, 30-34, partner, children,
quartile 4
Male, 30-34,
partner, children,
quartile 1
income
Probability (%)
27.6
income
9.7
income
24.3
income
60.8
income
50.2
income
97.2
income
96.5
income
99.8
income
99.6
15
3.3 Housing Tenure
The model determines the probability that a household with any of the characteristics
given in Table 6 will be an owner-occupier, private or social renter. Although the socalled Buy-to-Let market has added only 1.5 percentage points to the national housing
stock (4 percentage points in London), the importance of the rise in the sector is that it
provides a realistic alternative to owner-occupation for certain household types, namely
young, relatively high income groups, who require a high degree of mobility.
Consequently, the responsiveness to changes in the relative costs of the different tenures
might be expected to be high.
However, there are constraints on tenure choice. The most important is access to credit.
Since the early nineties the average age of entry to home-ownership has fallen sharply.
Although a number of factors have contributed to the decline, including the
improvements in the private rental sector and changes in the income distribution, the
increase in the required deposit stands out. For first-time buyers, the average deposit as a
percentage of the purchase price reached a maximum of 23% in 2003, compared with an
average of 12.5% between 1988 and 1997. Consequently, even if relative tenure costs
favour ownership, an inability to raise the required deposit is expected to extend the
period in the private rented sector. Note, however, that credit restrictions generally delay
ownership rather than permanently change tenure choice. Households have to save longer
to attain ownership.
Although economic factors are central to the determination of the tenure distribution,
Section 2 noted that official household projections indicate only a small percentage of the
increase in households will be amongst the under 35s – the main demographic group that
is attracted to private renting. Therefore, on the basis of demographic factors alone, a
drift away from private renting might be expected. Of course, this could be offset by
changes in relative prices and the credit restrictions discussed above.
A second constraint arises from housing supply and this is particularly important in the
social sector. Even if households would prefer to live in social housing, shortages in the
16
sector and the administrative allocation system mean that first choices cannot necessarily
be met. In the model, constraints of this form mean that households who do not meet the
eligibility criteria are forced into the low-cost part of the private rented sector.
Constraints bite more strongly in the southern regions than in the North or Midlands,
where areas of low demand exist. Low income households with children are consequently
more likely to be housed in social housing.
As in the household formation equations, the probabilities depend on tenure in the
previous year. Once again there is a high degree of persistence. Consequently, for two
households with identical characteristics, a household who was an owner last year will
have a higher probability of being an owner in the current year. As an illustration, Table
8a sets out, for a sample of household types, the estimated ownership probabilities in the
South East. The second column indicates that previous owners are highly likely to
maintain their status irrespective of demographic profile. But the probability that a
household who was previously a renter becomes an owner is fairly sensitive to income.
Table 8b conducts a similar analysis for renting. The table highlights the fact that social
renting falls sharply with income.
Table 8a. Ownership Probabilities for Previous Renters and Previous Owners (South East,
2003)
Female Head, Aged 30-34, Single, No Children
Previous Owner
0.936
Income Quartile 2
0.961
Income Quartile 4
Previous Renter
0.023
0.040
Male Head, Aged 35-39, Partner, With Children
Previous Owner
0.982
Income Quartile 2
0.991
Income Quartile 4
Previous Renter
0.078
0.120
Table 8b. Social Renting Probabilities for Previous Social Renters and Non-Previous Social
Renters (South East, 2003)
Female Head, Aged 30-34, Single, No Children
Previous Social Renter
0.899
Income Quartile 1
0.473
Income Quartile 4
Not Previous Social Renter
0.167
0.010
Male Head, Aged 35-39, Partner, With Children
Previous Social Renter
0.980
Income Quartile 1
0.763
Income Quartile 4
Not Previous Social Renter
0.428
0.064
17
3.4 Inter-Regional Migration
The inter-regional migration equations are estimated at the aggregate level on time-series
data. Therefore, the model does not consider the probability that a household with given
characteristics will move between regions. The variables included as determinants of
migration flows are conventional and cover those normally found to be important in the
literature. These include relative house prices, housing availability, relative
unemployment and earnings. Therefore, a combination of housing and labour market
influences are taken into account. Here, the emphasis is on the housing market.
As shown in the simulations below, migration flows limit the scope for affordability
targeting. Suppose, for example, that the supply of housing was increased in the South
East alone in order to improve affordability. As relative prices begin to fall in the South
East, out-migration from other regions would be expected to increase, particularly out of
London, worsening the current long-run trend towards the South East. Therefore, the
initial improvement in South East affordability would be offset. The problem becomes
more evident, the finer the spatial scale of targeting. At the local authority level, the
population flows are likely to be so great that it becomes impossible to meet a target,
since local authorities lie within wider travel-to-work areas.
4. Household Formation Under Market Outcomes
The official 2004-based household projections set out in Table 1 are trend based and may
suggest housing shortages could occur under current RPG figures. Under a market-based
solution, increases in housing costs would be expected to restore equilibrium by reducing
household formation, although, as noted earlier, this is likely to generate wider social
costs.
This section compares possible market solutions for household formation with the official
projections. However, it is convenient to begin with projections from the model in Figure
1 under an assumption that real housing costs remain approximately constant until 2029.
But, from the analysis in equations (2)-(4) this is inconsistent with the expected house
18
price path. To see this, assume long-run growth of real earnings of 2.5% pa. Over the
period 2007 to 2029, therefore, earnings would grow by approximately 75%. From
equation (4), if ∆r = 0 and ∆ln(HH) = ∆ln(Hs), then ∆ln(PH/Y) = 0.75, where ∆ = change
between 2007 and 2029. In other words, even if construction rises in line with the number
of households, affordability would be expected to worsen by 75% along the steady state
growth path. Since, for England as a whole, (PH/Y) = 7.0 in 2006, affordability might
worsen to more than 12 by 2029. In fact, due to short-run dynamics, in the simulations
below, the ratio is 11 in 2029.
Although there are considerable margins of error in any long-run projections, under an
assumption of constant real costs, the net flows from the model turn out to be similar to
official projections, although generally the model produces lower values, with the main
deviations occurring in London and the North West. But, to stress the point, they are
based on an assumption that costs are constant in real terms. Unconstrained, the model
would project increasing long-run real housing costs and, consequently, lower levels of
household formation. Moreover, the values in Table 9 imply housing shortages, since
house prices are not market clearing.
Table 9. Analysis of household Flows: Official Projections and Market Outcomes (000s)
(constant real housing costs)
GL
SE
E
SW
EM
WM
YH
NW
NE
England
20042029
Official
966
843
650
657
497
451
581
612
180
20042029
Market
854
827
631
606
470
423
579
532
173
20042029
Difference
-112
-16
-19
-51
-27
-28
-2
-80
-7
5437
5095
-342
By contrast, Table 10 shows similar information assuming market clearing, although the
table concentrates on the stock of households (rather than net flows) in the final year of
19
the official projections. The projections are, however, based on a particularly important
assumption. As in Section 3.1, rents – both private and social are assumed to rise in line
with owner-occupier housing costs. Since affordability rises sharply, unsurprisingly,
relatively few new households can afford to form.
By 2029 the difference between official and market projections is 4 million or 15% and
the differences are concentrated on the southern regions, where the affordability problems
are worst. In other words, the housing shortages are most severe in these regions and,
therefore, in a market driven system, the households cannot form. The figures imply a
rise in the average household size in the South East from 2.35 in 2004 to 2.45 in 2029.
Official projections suggest a continuation in the decline in average household size to
2.12.
Table 10. The Stock of Households in 2029: Official Projections and Market Outcomes
(000s)
(market clearing housing costs)
Official
NE
NW
YH
EM
WM
E
GL
SE
SW
England
Market
Clearing
Difference
1275
3507
2703
2296
2657
2954
4078
4211
2817
1221
3298
2562
2224
2564
2544
3102
3757
2419
54
209
141
72
93
410
976
454
398
26497
22470
4027
Arguably, Table 10 reflects an unrealistic worst case scenario, but it illustrates (i) how
strongly the housing system relies on social support and (ii) the dangers of assuming that
past trends will continue. It was noted in Table 3 that past official projections overestimated the outturn during the nineties. It may be the case that current official
projections will again over-estimate the outcome, because of a failure to allow
sufficiently for affordability changes.
20
Nevertheless, despite these warnings, we would not want to argue that the model results
in Table 10 are the most likely outcome – the model simply provides a framework for
thought, which allows for affordability changes, although household formation could well
be below official estimates. The social implications of a rise in household size are likely
to be severe and would probably not meet the government commitment that “everyone
should have the opportunity of a decent home at a price they can afford, in a place in
which they want to live and work”. Furthermore, the results rely on social housing costs
rising in line with house prices, which, in turn, imply an unrealistic excess supply of
social homes. In practice, the majority of social tenants and low income private renters
would be receiving housing benefits, so the costs they face would be lower or zero and
household formation would be correspondingly higher than in Table 10. But this, of
course, has implications for the benefits bill.
5. Policy Questions
In this section a selection of current policy issues is considered. First, the paper discusses
the difficulties of raising homeownership rates, given that home-ownership rates in
England have barely changed from 70% over the past five years. The discussion is tied up
with the more general question of home-ownership sustainability. Second, the paper
examines the impact of increasing housing supply, a recommendation of the Barker
Review. We attempt to dispel some popular fallacies arising from the Report. For
example, the Report does not necessarily imply large amounts of new building on
Greenfield sites, nor the break up of urban communities. Nor does the Review imply the
demolition of Victorian terraces. The criticism that there are already enough dwellings to
meet demand is also found not to be valid. Third, the section discusses the feasibility of
achieving affordability targets.
It is useful to begin with an examination of expected tenure shares implied by Tables 9
(constant real housing costs) and Table 10 (worsening affordability). To recap, despite
the uncertainties surrounding long-run projections, Table 9 suggests household numbers
21
similar to official projections, whereas by 2029 – assuming rents rise in line with house
prices – households are approximately 4 million lower than official estimates in Table 10.
The 4 million represent households who fail to form because they cannot meet the market
rent. However, as we have stressed, there are strong social reasons why these households
need to form, given the role of housing as a merit good. In the following, therefore, we
assume that the 4 million do form, but at subsidised housing costs in the rental sectors.
Formally, the 4 million are allocated, first, to the social sector up to the point at which
shortages occur (allowing for vacancies) and the remainder are allocated to the private
rented sector. In the bottom half of Table 11, therefore, housing “need”, as represented
by official household projections, is met (rather than effective demand), but through the
rental sectors. The two halves of the table, therefore, are based on similar household
numbers. The tenure shares in Table 11, concentrate on an intermediate year, 2015, since
the aspirations to raise ownership rates are well before 2029.
Table 11. Tenure Shares
2001
2004
2015
Owners
Social Renters
Private Renters
70.5
19.9
9.7
70.0
19.9
9.7
71.9
18.7
11.3
Owners
Social Renters
Private Renters
70.5
19.9
9.7
70.0
18.7
11.3
70.4
17.5
12.1
England
Constant
Real Costs
Worsening
Affordability
Under constant real housing costs, for England as a whole in 2015, the model implies that
the ownership rate might be approximately 72%3 compared with 70% currently.
Therefore, even without a worsening of affordability, there is likely to be work to do to
extend ownership towards 75%, as mentioned in the Government’s response to the
Barker Review of Housing Supply, perhaps through a combination of Right to Buy Sales,
shared ownership or a relaxation of credit conditions. But, under worsening affordability,
Table 11 implies that home-ownership might change little from current rates.
The
difference between the two scenarios arises because low income groups can no longer
3
This assumes no further Right to Buy sales.
22
afford to become households in the market sector and are forced into the social sector. If
official household projections are to occur, therefore, the difference between modeldetermined household numbers and the official projections are in the rental sectors.
Sustainable home ownership can have a number of meanings (see Meen 2006) and the
term is often used loosely. Here, we mean the rate of ownership that can be maintained
over time without sharp fluctuations in house prices, output, arrears and possessions. An
implication is that permanent increases in ownership can be achieved only by increases in
supply; this, in turn, can be brought about by (i) new building, conversions, renovations,
change in use from commercial or retail properties, (ii) transfers between tenures, notably
a high degree of substitution in supply between renting and owning. Arguably, (ii) has
increased in recent years with the advent of the Buy-to-Let market. The expansion in
ownership through supply measures has an analogy in the labour market literature, one
strand of which suggests that unemployment can only be reduced permanently through
supply-side reform. Under this view of the world, government attempts to expand
demand leads, in the long-run, to inflation rather than higher output or lower
unemployment. It follows that housing demand subsidies or attempts to reduce arrears
and possessions below the “natural” rate add to house price inflation rather than raising
the long-run ownership rate. Furthermore, shared-ownership schemes or reductions in
deposit requirements bring forward the timing of entry into ownership rather than raising
the long-run level.
From equation (4), stabilisation of affordability requires the supply of housing services to
rise faster than the number of households. As noted above, the simulations in Table 10
imply that affordability worsens sharply between 2006 and 2029. Although too much
emphasis should, perhaps, not be placed on the precise numbers, since any projections
including house prices are subject to large margins of error, the implications of the
analysis are clear. Therefore, the effects of increasing the supply of housing services are
considered next. Although the supply of services could be expanded through renovation
of the existing stock, the simulations make net additions to the stock. These could be
23
through a combination of new building, conversions or changes in use. The annualised
RPG targets in Table 4 provide the baseline, on top of which a further 50,000 dwellings
per annum are added, spread proportionately over the four southern regions, where
shortages are greatest. However, before conducting the simulations, the expected
outcomes can be seen analytically. Rewriting equation (4) as (5):
∆ ln( PH / Y ) = b0 + b1∆ ln(Y ) − b2 ∆ ln( H s / HH ) + ε 2
(5)
b0 = (a0 / a2 )
b1 = (a1 / a2 − 1) = 1.0
b2 = (1 / a2 ) = 2.0
∆r = 0
The coefficient values arise directly from the earlier analysis. Given the growth in real
earnings, stabilisation of affordability requires ∆ ln( H s ) > ∆ ln( HH ) . Again if real
earnings rise by 2.5% per annum between 2007 and 2029, the housing stock would have
to rise by 1.25% per annum more than the increase in households. But, from Table 6, the
number of households in any region is negatively related to affordability (6):
∆ ln( HH ) = −c1∆ ln( PH / Y ) + ε 3
(6)
Therefore, as the housing supply improves affordability, further households are induced either
through new household formation within the region or through inter-regional migration. Note that
if ∆ ln( HH ) = 1.0∆ ln( H s ) , then increases in housing supply can never stabilise affordability.
This reverses the traditional view that increases in construction should match the number of
households.
The Introduction pointed to 3 key elasticities that determine the outcomes. The first has already
been discussed, whereas equation (6) brings together the second and third. The model outcomes
are summarised in Table 12 for 2015. The first column shows the improvement in affordability in
24
each of the four regions; the second the increase in households operating in the market sector4,
the third the cumulative change in migration flows between 2007 and 2015 and the fourth the
cumulative increase in construction over this period. Although fairly modest, at approximately
0.7 points, increases in construction do improve affordability. But, more controversially,
increases in construction are not matched by increases in the number of households. In London,
for example, although construction increases by 128,000 over the period, households rise by less
than 50,000 or 40%. The ratio of households to construction is even lower for the other regions
and is approximately 20% over the four regions. We have already seen that this is a necessary
condition for improvements in affordability. But it gives rise to the question of who will live in
the extra homes, related to issues of vacancies and demolitions. The higher responsiveness in
London is because, as noted earlier, London has a younger population structure and, therefore, a
smaller percentage have already formed separate households. In a region where a high proportion
of the population have already formed households, a further improvement in affordability has
limited effect. An alternative characterisation of the results is to estimate the elasticity of
household formation with respect to house prices; in London the elasticity is -0.2, but less than 0.1 in the other regions.
Also, from Table 12, the migration flows are limited; this is because the increase in
construction is “balanced” across the southern regions, so the relative price effects are
small. The migration figures are not directly comparable with those in the previous
column, which refer to households rather than population.
Table 13 repeats the
simulation, but construction is only increased in the South East. In this case, migration
and the increase in households are greater, but the improvement in affordability is more
modest. More generally, this suggests that spatial targeting at fine levels is difficult.
Within travel-to-work areas, for example, population flows offset attempts to improve
affordability through increases in construction.
Table 12. Increases in Housing Construction: Balanced Across the Southern Regions
(changes from base scenario, 2015)
London
South East
East
Affordability
(% points)
-0.71
-0.66
-0.72
Households
(000s)
49
19
9
4
Migration
(000s, 2007-15)
21
-5
-12
Construction
(000s, 2007-15)
128
131
97
Total households are still constrained to official projections, but at improved affordability a greater
percentage can operate in the market sector.
25
South West
-0.77
15
4
94
Table 13. Increases in Housing Construction: South East Only
(changes from base scenario, 2015)
South East
Affordability
(% points)
0.51
Households
(000s)
48
Migration
(000s, 2007-15)
55
Construction
(000s, 2007-15)
131
The Barker recommendations have, in some quarters, been interpreted as a call for large
increases in new building on Greenfield sites and for the break up of existing urban
communities through demolitions (see Power and Houghton 2007). There has also been
the suggestion that there are already sufficient dwellings because of current large
numbers of vacancies. In fact, the Review said relatively little about the spatial
distribution of any extra dwellings and none of the analysis above necessarily implies a
preference for Greenfield new homes over the renovation or conversion of the existing
stock within cities. To see this note, first, that the simulations in Tables 12 and 13 are in
terms of net additions to the stock. Although, from Table 5, a high percentage of net
additions have historically been through new construction, conversions and changes in
use also make a significant contribution and could potentially be extended. The main
requirement, as this paper stresses, is the need to increase the volume of housing services
to meet the increased demands over time of existing households, rather than increasing
the number of units.
Second, there is a suggestion that the Barker proposals involve high levels of demolitions
of older (predominantly Victorian terrace) properties. As Table 5 showed demolitions in
England are low as a percentage of the housing stock (and have been so since the ending
of the large slum clearance programmes between the fifties and seventies), but recent
attempts to increase demolitions in the Pathfinder areas have been unpopular. In fact,
Meen et al (2006) argue that, although there is a case for increasing demolitions in order
to improve the quality of the housing stock (and consequently raise the quantity of
housing services), older Victorian terraces are not the most likely candidates. Hedonic
regressions indicate the properties that are least valued are those built in the 1950s and
26
1960s. By contrast, Victorian terraces are typically highly valued since only the better
parts of this vintage of the stock have survived to the current day. But the important point
is that any demolitions should be based on market valuations and concepts of economic
obsolescence rather than age per se. Since the older parts of the stock have an urban bias,
it is not the case that demolitions necessarily involve inner city destruction.
A further criticism is that high levels of new building are not required because England
already has a high level of vacant properties that could be brought back into use.
Although international comparisons are difficult, Evans and Hartwich (2005) indicate
low vacancies in England. A priori, this would be expected because real house prices are
high by international standards and the opportunity cost of holding dwellings empty is
high. HIP return data suggest that private sector vacant dwellings in 2003 were
approximately 3.5% of the housing stock, of which 1.8% had been vacant for more than
six months. In a freely operating housing market, a certain level of vacancies is necessary
as part of the transactions process and, therefore, the scope for reducing vacancies further
may be limited. Indeed, they may need to be higher if the opportunity cost declines under
affordability target.
Finally, it has been argued that large increases in the number of homes would exceed
official household projections. Furthermore, Table 12 implies that few of the increase in
homes go to new households. However, once again, the key concept is the increase in
housing services, rather than the number of units. Nevertheless, it is likely that the
increase in services would have to be met partly from an increase in units and, given that
equation (1) must hold, the point is valid. However, equation (1) also stresses that
equilibrium can be restored not only by an increase in households, but also through a rise
in second homes, vacancies, or demolitions. The fact that each of these has, historically,
been fairly small is a reflection of past housing supply shortages and each would be
expected to be part of the adjustment in a market system. This process would involve a
filtering of the housing stock down through income groups. As already discussed, the
lowest valued parts of the housing stock would be demolished, but to stress the point, this
is not necessarily older units within cities.
27
6. Conclusions
Taking into account affordability requires changes to the traditional planning approach,
where housing net additions are closely allied to expected, trend-based household
formation. In particular, estimates have to take into account the effects of changes in
affordability on the demand for housing services by existing households and the fact that
targets are likely to change the historical trend formation rates for new households.
Although the externalities associated with housing and government objectives for decent
homes continue to mean that housing cannot be entirely treated as a pure marketdetermined outcome, affordability targets imply that greater account has to be taken of
market processes than in the past as part of the planning process.
The paper shows that matching household formation and the number of new homes in
terms of units is not sufficient to stabilise affordability, measured by the ratio of house
prices to earnings. The provision of household services would have to rise at a faster rate.
Alternatively, if affordability worsens, then household formation would be lower than
implied by official projections unless low income households continue to be subsidised in
the social or private rental sectors. It is noticeable that past projections of household
formation during the nineties over-estimated the increase, possibly due to worsening
affordability, whereas projections for the future have been revised upwards. Worsening
affordability also implies that it becomes difficult to raise significantly rates of homeownership. Increases in shared ownership and/or relaxation of credit conditions,
typically, change the timing of entry to home-ownership rather than the long-run
sustainable level.
The paper also suggests that the Barker proposals to increase housing to improve
affordability do not necessarily imply covering large areas of the Green Belt, nor the
demolition of good-quality Victorian homes, nor the decline of urban centres. The key is
to increase the quality of the housing stock to raise the quantity of housing services. This
28
can occur through a combination of new building, renovation, conversions and changes in
use.
References:
Andrew, M. and G. Meen (2003). “Housing Transactions and the Changing Decisions of
Young Households in Britain: The Microeconomic Evidence”. Real Estate
Economics, 31(1): 117-138.
Barker, K. (2004). Review of Housing Supply: Final Report (London: HMSO)
Blackaby, D.H., and D.N. Manning. (1992). “Regional Earnings and Unemployment – A
Simultaneous Approach.” Oxford Bulletin of Economics and Statistics. 54: 481-502.
Bover, O., J. Muellbauer., and A. Murphy. (1989). “Housing, Wages and UK Labour
Markets.” Oxford Bulletin of Economics and Statistics 51(2): 97-136.
Bramley, G., M. Munro., and S. Lancaster. (1997). The Economic Determinants of
Household Formation: A Literature Review. Report for the Department of the
Environment, Transport and the Regions.
Cameron, G, J. Muellbauer and A. Murphy (2006), “Was There a British House Price
Bubble? Evidence from a Regional Panel”, Department of Economics, University of
Oxford mimeo.
DiPasquale, D. and E.L. Glaeser (1999), “Incentives and Social Capital: Are
Homeowners Better Citizens? Journal of Urban Economics, 45: 354-384.
Evans, A.W. and O. Hartwich (2005). Unaffordable Housing: Fables and Myths, Policy
Exchange. London.
Meen G, (2005), “On the Economics of the Barker Review of Housing Supply”, Housing
Studies 20(6): 949-971.
Meen G.P (and 15 others) (2005). Affordability: Implications for Housing Supply,
Technical Appendix. ODPM. London.
Meen, G.P., and M. Andrew. (1998a). “On the Aggregate Housing Market Implications
of Labour Market Change.” Scottish Journal of Political Economy 45(4): 393-419.
Meen, G.P., and M. Andrew. (1998b) Modelling Regional House Prices: A Review of the
Literature. Report prepared for the Department of the Environment, Transport and the
Regions. The University of Reading.
Meen, G.P, J. Meen, and C. Nygaard (206), “A Tale of Two Victorian Cities in the 21st
Century”, paper presented to the ENHR meeting, Heriot-Watt University.
29
Muellbauer, J., and A. Murphy. (1997). “ Booms and Busts in the UK Housing Market.”
Economic Journal 107(445): 1701-1727.
Power, A and J. Houghton (2007), Jigsaw Cities: Big Places, Small Spaces. The Policy
Press, Bristol.
Rohe, W.M., S. Zandt and G. McCarthy (2002), “Home Ownership and Access to
Opportunity”, Housing Studies, 17(1): 51-62.
30