House price inflation

Question 1

How is ONS index constructed?

Answer 1

• Varied over time - most recent from review process 2010-2015
• HPI constructed each month by hedonic regression using that months transactions – 2 months lag
• Around 100,000 houses sold every month
• Selection bias as some house sold much more than others – e.g., some stay in a house for their whole life
• Transactions mix may alter avg p – in some months avg p may be higher due to mix of transactions e.g., if there are a lot of high val homes sold
• Hedonic method controls for quality of houses – ONS index is a constant quality index
• Hedonic model represents price transacted from PPD Pi as function of properties of the house itself
• Coefficients give proportional effect on transaction p – easy to interpret

Question 2

Where do we get data - price paid data?

Answer 2

• Source – Land registry – England and Wales
• Standard Price Paid Data – full market p – inc all purchases by private indiv whether this inc mortgage element or complete cash – will inc buy to let if in cash – in May 2017, SPPD inc 78,846 transactions
• Additional Price Paid Data – first in oct 2013 – inc few others – transfers to non private indiv (e.g., a company), transfers under power of sale (repossessions) may 2017 inc 14,986 transactions
• Current HPI uses SPPD plus buy to lets from additional PPD
• Available info to public
• Inc location, type of property

Question 3

Where do we get data - VOA dataset?

Answer 3

• Provides property characteristic
• Organisation that deals w council tax, business rates, collects rental data
• Not public
• E.g, no habitable rooms, type of property, floor space
• Detached, Semi-detached, terrace, flat
• Dummy variables, other than for floor space
• Doesn’t inc land area, garage, age of building etc.

Question 4

Where do we get data - socioeconomic variables?

Answer 4

• Acorn – powerful consumer classification segmenting UK pop
• Provides precise info and understanding of diff types of people
• Post codes classified
• Segments postcodes and neighbourhoods into 6 categories, 18 groups and 62 types – three of which are non private HH
• ONS uses these 18 groups
• Able to provide more refined targeting
• Marketing tool, but used for HPI

Question 5

Hedonic regression?

Answer 5

• DV –
• Transaction price – where is full market p, not right to buy etc.
• Explanatory variables – all dummy except for floor space –
• Location – Eng 325, Scot 32, Wales 22
• House characteristics –
• New build, D S T F – higher cost for new build
• Socioeconomic – using Acorn
• For each house you will have a predicted val Phat i and regression coefficients bhat chat
• HPI is geometric mean
• Log-linearity means val of HPI dep only on avg val of explanatory variables, not distrib across houses
• Weights Xbar Ybar – these multiply the estimated val to give index
• Weights for each characteristic updated annually and are the avg found in last yrs total transaction
• They are in effect the shares of each characteristic
• Weights are fairly stable
• Published HPI is then HPI = exp[lnHPI]
• ONS avg house price is a geometric measure
• Coefficients vary month to month but not by a huge amount
• Coefficients for Acorn generally slope down to reflect ‘better’ areas at the top

Question 6

Stock verus flow?

Answer 6

• Weights are for the flow of housing transactions – small proportion of the total stock
• England around 23.7m dwellings, 2016 around 4.2% of total stock sold
• The houses sold are not the same as the stock – some types of house sell more frequently than others
• Main bias is the new builds as shown
• About under/over representation
• The flow is mainly OOH so rep the higher Acorn groups
• Social housing sales are small – often rep poorer HH groups – stock inc more poor less advantaged areas
• Stocks HPI – level is lower, but inflation HPI rate is similar

Question 7

Alternative measures of HPI?

Answer 7

• Differ by data used – house price transactions PPD, mortgage applications approved, advertised prices
• Mortgage approvals can inc a lot more info ab house – garage, no toilets, method of construction
• PPD and mortgage applications – backward looking – p agreed and application some time before actual sale
• Can be some lag
• Advertised p – forward looking – may differ from eventual transaction p

Question 8

Differing measures of HPI?

Answer 8

• If using mortgage applications – will have much lower no transactions due to lack of access to data
• Mix adjustment – dividing the data into cells, seeing how the mix adjusts over time – cells weighted according to past transactions/stock
• Most characteristics similar across indices but differ in detail used
• Differ by mean
• Geometric -ONS, Nationwide, Halifac
• Arithmetic (mix adjusted) – Rightmove, LSL
• Geometric avg lower than arithmetic

Question 9

Timeliness of measures?

Answer 9

• Rightmove is timeliest, estimates during reference period
• Nationwide and Halifax – one week after reference – Sep index published in early October
• ONS least timely – publishes HPI 6 weeks after end of month

Question 10

Current situation?

Answer 10

• ONS – sep 2023
• UK avg house p decreased by 0.1% over the year, down from 0.8% in August
• Rightmove – Nov 2023 – avg new seller asking p dropped by 1.7% as Christmas approaches and adopting more price realism to attract buyer
   Though the transition from the frenzied pandemic market back towards more normal activity levels has been slow, key indicators point to a year that so far has been better than many predicted following the turbulent end to 2022