Local Housing Supply & the Impact of History and Geography

April 24, 2023 © University of Reading 2006 www.reading.ac.uk

Local Housing Supply & the Impact of History and Geography Geoff MeenAndi Nygaard

2To put your footer here go to View > Header and Footer

Context • Work for DCLG on housing supply elasticities at

different spatial scales: international to national to local to firm level. Different scales provide different insights. This paper looks at local level

• 2 issues:(i) Impact of planning policy(ii) history/geography property rights.• Once an area is developed, the transaction (and

building costs) rise. Therefore, expect less development on brownfield sites adds to spatial fixity of the existing stock. Low demolition rates.

• But policy works in the opposite direction.


Sub-plot • Attempts to test conventional wisdoms: Weak

housing price elasticities of supply are due to the planning system.

• Barker Review of Housing Supply policies to increase housing supply, through loosening the planning system.

• But…1. Could weak supply elasticities be due to other

factors? If so, basis of policy is undermined.2. Is it really true that supply elasticities are lower in

UK than internationally or are there methodological differences between international studies? Led to international comparisons (but not topic for today).


Sub-plot (2) • At the local level, one reason for low

supply elasticities could be the historical pattern of land use (property rights built up over centuries) and physical geography (hard to build on water, marshes, mountains).

• Therefore low supply elasticities in particular areas could simply reflect differences in land-use patterns.

• If so, unfair to impose common targets on local authorities.


Study Areas • Thames Gateway versus Thames Valley.• TG is area for special policy action – pockets

of high poverty, relatively poor transport links, government targets for increasing housing supply to reduce South’s housing pressures. Includes Olympic park.

• TV is one of richest areas in country, good transport links etc.

• Analysis is at MSOA level (approx. 3,000 dwellings/households). 381 in TV, 210 in TG.


Land Use Table 1. Land Use in the Thames Valley and Thames Gateway (%)

Land Use Thames Valley Thames Gateway

Domestic buildings & gardens 6.44 11.99

Non-domestic buildings 0.62 1.89

Roads 2.24 4.70

Paths 0.11 0.31

Rail 0.12 0.35

Green space 87.59 58.60

Water 1.42 17.76

Other 1.46 4.39

Total 100 100

Source. ONS Neighbourhood Statistics (Generalised Land Use 2005)


Housing Change


Price Gradients

Figure 3. Price-Distance Functions – Thames Gateway

Price distance (km)

5060

7080

90100

110120

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150

Price ratio 06 Price ratio 05 Price ratio 04 Price ratio 03

Figure4. Price-Distance Functions – Thames Valley

Price-distance TG to Central London

30 35 40 45 50 55 60 65 70 75 80

Distance 8.8 13.8 18.8 23.8 28.8 33.8 43.8 48.8 53.8 58.8 63.8 68.8 73.8

Kilometers

Percent of Centre

Price 2003 Price 2004 Price 2005 Price 2006


Spatial Fixity • One view is that major structural change in any

area only takes place infrequently, so that we shouldn’t expect big responses to modest changes in prices.

(i) exogenous innovations: examples are wars, acts of terrorism, acts of God.

(ii) Policy innovations: these include major infrastructure changes, e.g. new

road networks, new social housing estates, slum clearance and major

regeneration schemes such as the Olympics.

(iii) endogenous innovations: migration is the most notable.

(iv) Technology innovations: for example, the Industrial Revolution, powered

flight, motorised transport.


A basic model

iiiie

iiii HWHHPXPHXH 3431211 .)()( (1)

iiiii PHPHXHHP 432210 )*( (2)

*.* 3221 iiii PHWXHPH (3)

H = housing stock PH = house prices X1 = vector of geographical, historical and policy variables that affect the supply elasticities X2 = vector of locational variables affecting housing demand, e.g. distance to London and

employment centres, socio-economic status W = spatial weights matrix (i) = MSOA (e) = denotes expected value (*) = denotes equilibrium value (ε) = denotes an error term (.) = denotes a time derivative.


Or ….

CZBYYA (4)

A=

1

)()1(

1

124

XW

B =

)(

)(

33313

113

WX

C =

230

00

Y =

HPH

Z =

2

1X

=

4

3


An operational model: Prices

iiii

ii

iiiii

NONDOMDOMDIST

IMDROUTINEINTERMHMANHHHSTRAVELPH

51098

76

54321

)()()ln(

)ln()()()()/ln()ln()ln(

(5)

PH = median house price TRAVEL = average travel to work distance or distance to secondary labour market (HS/HH) = number of dwellings/ number of households HMAN = percentage of population in higher managerial and professional occupations INTERM = percentage of population in lower managerial, professional and intermediate occupations. ROUTINE = percentage of population in routine occupations. IMD = index of multiple deprivation DIST = distance to Central London DOM = percentage of area devoted to domestic buildings NONDOM = percentage of area devoted to non-domestic buildings


An operational model: Net Additions

ijij

j

kikik

kkik

kiii

iiiii

LAD

PHSUBSUBHSWWATERGREEN

GARDENPATHHPPHHS

6

18

1

7

1

7

107/04765

4306/032,031007/04

)(

)ln(*)()()ln(.)()(

)()()ˆln()ln()ln(

(6)

ijij

jkikik

kkik

ki

iiiiii

iiiii

LADPHSUBSUBHSW

HPWATERHPGREENHPGARDEN

HPPATHHPPHHS

7

18

1

7

1

7

107/047

06/03606/03506/034

06/03306/032,031007/04

)()ln(*)()()ln(.

)ˆln(*)()ˆln(*)()ˆln(*)(

)ˆln(*)()ˆln()ln()ln(

(7) HS = housing stock (public and private)

HP ˆ = expected median house price, derived from (9)

03PH = median house price in 2003 PATH = proportion of land area devoted to paths GARDEN = proportion of land area devoted to domestic gardens GREEN = proportion of land area devoted to green space WATER = proportion of land area devoted to water SUB = housing sub-market dummy variable LAD = local authority dummy variable

07/04 = change in the variable between 2004 and 2007

06/03 = change in the variable between 2003 and 2006.


Results: PricesTable 3. Modelling Median House Prices in the MSOAs (2003 & 2006). Dependent variable = ln(PH) Variable 2003 2006 OLS constant 15.874 (58.8) 15.547 (58.4) DTV -1.631 (7.1) -1.332 (6.0) HMAN 0.019 (7.9) 0.020 (8.1) INTERMED -0.023 (14.3) -0.024 (15.3) ROUTINE -0.046 (12.3) -0.048 (13.1) ln (IMD) -0.233 (8.6) -0.168 (6.3) ln(DIST) -0.116 (6.9) -0.105 (6.3) DOM -0.010 (6.9) -0.010 (7.2) NONDOM -0.006 (2.1) -0.005 (2.0) DTG*ln(TRAVEL) -0.191 (3.0) -0.111 (1.8) DTV*ln(TRAVEL) 0.028 (2.9) 0.039 (4.3) DTV*ln(IMD) 0.176 (6.3) 0.116 (4.3) DTV* (NONDOM) -0.014 (3.6) -0.014 (3.9) R2(adj) 0.83 0.83 Eqn. St. Error 0.133 0.131 N 591 591 t-values in brackets. DTV = dummy variable for the Thames Valley MSOAs; DTG = dummy variable for Thames Gateway MSOAs.


Results: Net AdditionsTable 5. The Growth in the Housing Stock. Dependent variable = )ln(07/04 HS Variable Coefficient Coefficient Coefficient OLS 2SLS constant -0.810 (4.3) -0.983 (4.9) -0.979 (3.5) DTV 0.725 (3.6) 0.834 (3.9) 0.807 (2.8) ln(PH)03 0.068 (4.6) 0.074 (4.8) 0.072 (3.2) PATH -0.0081 (2.8) - - GARDEN -0.0018 (8.1) - - GREEN -0.0011 (8.2) - - WATER -0.0008 (4.2) - -

)ˆln(06/03 HP 0.478 (4.4) 0.988 (8.2) 0.954 (6.4)

DTV*ln(PH)03 -0.057 (3.4) -0.066 (3.6) -0.063 (2.6) DTG* )ˆln(06/03 HP *PATH - -0.027 (1.4) -0.026 (1.4)

DTG* )ˆln(06/03 HP *GARDEN - -0.010 (6.5) -0.010 (3.6)

DTG* )ˆln(06/03 HP *GREEN - -0.006 (7.0) -0.006 (4.0)

DTG* )ˆln(06/03 HP *WATER - -0.004 (3.8) -0.004 (2.1)

DTV* )ˆln(06/03 HP *PATH - -0.062 (2.5) -0.061 (2.6)

DTV* )ˆln(06/03 HP *GARDEN - -0.010 (5.5) -0.010 (3.9)

DTV* )ˆln(06/03 HP *GREEN - -0.006 (5.7) -0.006 (4.2)

DTV* )ˆln(06/03 HP *WATER - -0.006 (2.6) -0.006 (3.4)

)ln(. 07/04 HSW - - 0.785 (0.9)

R2(adj) 0.599 0.600 0.599 Eqn. St. Error 0.028 0.028 0.028 N 591 591 591 Equations include dummy variables for the local authorities and also dummies for 7 MSOAs exhibiting outlying errors. t-values in brackets.


Analysis (1)

The UK government announced a target for England in 1998 that at least 60% of new

homes should be built on previously developed land by 2008. As shown below, there is

little doubt that the policy affects aggregate elasticities.

But, given that the national policy is common to all areas, the estimated model can be

used to examine the extent to which variations in price elasticities across space reflect

existing land use patterns determined by history and geography. Price elasticities in some

local authorities may be weaker than in others, not because they are necessarily less

compliant with national policy objectives, but because they have inherited particular land

structures.


Analysis (2)

The interpretation of the interactive price elasticities is not straightforward. If an area has no green space, gardens, water or paths, i.e. the area is fully built up, then the elasticity in terms of price changes is approximately 1.0 in both areas (but not a conventional price elasticity). But, at the mean land use values across the two areas (PATH = 0.7%; GARDEN = 20.4%; GREEN = 50.5%; WATER = 3.7%), the price elasticity for the Thames Gateway is 0.44 and 0.41 for the Thames Valley. Therefore, (i) areas that are heavily built up have noticeably higher price elasticities (policy effect), but (ii) differences in the elasticities between the two areas are modest, once account has been taken of cross-area existing land use variations.


Analysis (3)

To illustrate, using the mean land use percentages within each area (rather than the joint

mean), the elasticity in the Thames Valley falls to 0.38, whereas that in the Thames

Gateway remains similar at 0.48.

Therefore, the differences in responses arise primarily because of a higher proportion of

built up areas (brownfield land) within the Thames Gateway. Note, however, that areas

that have particularly high percentages of gardens, which, of course, may accompany

development, have low elasticities (coefficient value = -0.01). It may be speculated that

this reflects the pressures against development imposed by existing residents or property

rights issues.


Analysis (4)

Nevertheless, even these area differences are modest, but variations in land use patterns

between the MSOAs within each area produce much bigger effects on the elasticities and

expected construction growth rates.

This is particularly noticeable in the Thames Gateway, where the estimated standard

deviation, across MSOAs, for the growth rate in the housing stock between 2004 and

2007 falls by approximately 40% on the assumption that land-use shares in each MSOA

are equal to the average. More details are presented in Figures 6 and 7, On this “flat

plain” assumption, development is more spatially evenly distributed.


Flat Plains


Analysis (5)Within the constraints of a national planning policy, the distribution of land types clearly

matters. In the absence of controls, lower elasticities would be expected on previously

developed land, because of the additional costs (including transactions costs). Here,

higher elasticities are found, with the exception of areas where gardens are a high

proportion of land use.

But, suppose that controls were lifted on greenfield development. The price elasticity of

supply cannot be derived under these conditions, since they are never observed, but a

minimum bound is given by the maximum elasticity across the MSOAs under current

policy (0.82). As expected, this occurs in a highly built up area (with few gardens), where

government planning controls are not expected to bind.

However, this is a minimum value and the elasticity on greenfield sites is likely to be

higher, because of lower development and transactions costs. Furthermore, the minimum

estimated elasticity in the model is 0.14. Therefore, the maximum elasticity is

approximately six times the minimum. Clearly, existing land use has a strong effect on

the price elasticity.


Conclusions

This paper considers the impact of existing land-use patterns on housing supply elasticities in local areas of England, under existing planning policies. The paper demonstrates that, despite common national planning policies, local supply responses to market pressures vary considerably, because of differences in historical land uses. Due to differences in historical land use and geography, the price elasticity in the least constrained area is approximately six times higher than the most constrained. Given the considerable local variations, a question arises whether it is reasonable to expect all local authorities to meet the same targets. But the conventional wisdom still seems to be correct – planning policies lower price elasticities.

Local Housing Supply & the Impact of History and Geography

Documents