The R

1
The Rôle of Price Expectations in the U.K.
Housing Market
2
Introduction

1. Forward-looking expectations play a crucial
   lead in the determination of house prices for
   builders, buyers, sellers, etc.
2. The majority of housing studies include only the
   expectations of the general price level, if they
   deal with expectations at all.
3. It is suggested in the theoretical literature
   that there is a strong link between housing
   expectations and the state of the economy. A
   factor in the latest recession. For example,
   expectations of falling house prices can reduce
   consumer spending.
4. The paper explains theoretically the formation of
   expectations and applies that to RICS survey data
   of house price expectations over the next three
   months, or more, in conjunction with both the
   Nationwide and Halifax actual house price
   indices.

3
The Data

1. There are several sources of actual data on house
   prices the Halifax, the Nationwide, the Land
   Registry and the Financial Times indices.
2. The mortgage lenders series provide the longest
   run of monthly, non-seasonal statistics, and
   therefore, are adopted in this empirical
   investigation. Also, the Land Registry series is
   behind the times because it lags actual
   transactions the lenders series by contrast,
   lead transactions.
3. The Figure on the next slide shows the
   fluctuations in the logarithmic growth rate of
   Halifax prices over the next three months, that
   is the logarithm at time (T4) minus the
   logarithm at time (T1). The Nationwide index
   would give similar results.

4
(No Transcript)
5
The Theoretical Model

1. The formation of forward-looking expectations can
   be modelled on the basis of bounded rationality
   and the interdependence of agents, described as a
   process of diffusion.
2. Those agents with the resources to form accurate
   expectations cheaply are a small number of
   chartered surveyors who possess the knowledge of
   the market and are part of an Institution that
   publishes the expectations in the form of either
   up, same or down .
3. This group is small relative to the majority,
   which means that the distribution of expectations
   will be initially slow, followed by a sudden
   increase as the majority of agents adapt to the
   change in predictions.
4. This sequence implies a nonlinear process of
   diffusion, which can be captured by the logistic
   function, represented by a S-shaped curve, shown
   in the next slide.

6
(No Transcript)
7
The Empirical Estimation of the Diffusion Model

1. A logistic model of expectation diffusion is
   estimated using the RICS data on future price
   trends. Given that the sames have changed
   considerably over the time period, it is
   necessary to normalise the data.
2. The ups and downs can be normalised so they
   sum to one (or hundred), by calculating adjusted
   variables. Then, either one of normalised ups
   or normalised downs can be used because they
   give perfectly symmetric results.
3. A technical difficulty, however, is that the
   normalised ups contain some zeros, and it was
   not possible to calculate the logarithms for
   them. To overcome this problem, the zeros were
   adjusted to 0.005 and the ones reduced to 0.995
   so that the logistic variable could be derived.

8
More on Empirical Estimation

1. The statistical analysis, which explains the
   logistic diffusion variable derived from the RICS
   survey using the adjusted ups, employed the
   Hendry methodology of general-to-specific
   analysis together with price data from the
   Halifax database over the period of 1999 M10 to
   2009 M9. This is followed with the substitute
   data set, the Nationwide, over the same period.
2. The significant variables in the equations are
   explained on the next slides.
3. Both empirical models are similar.

9
Halifax Model explaining RICS

• The dependent variable is the change in the
  log(au/(1-au)) from the RICS survey
• Explained by the forward-looking growth in the
  Halifax price index, logarithm T4 minus
  logarithm T1
• The dependent variable lagged 1, 5, 9 and 15
  months
• The Halifax growth lagged 16, 21 and 22 months
• One dummy variable for September in 2004
• The dynamics here are quite complicated,
  suggesting some form of error correction.

10
Nationwide Model explaining RICS

• The dependent variable is the change in
  log(au/(1-au)) from the RICS survey
• Explained by the forward-looking growth in the
  Nationwide price index, logarithm T4 minus
  logarithm T1, lagged one month
• The dependent variable lagged 5 and 12 months
• The Nationwide growth lagged 14 and 24 months
• Two dummy variables for September 2004 and
  December 2008.

11
Econometric Estimation of the Forecasting Models

1. The forecasting of future growth in house prices
   is causally quite different to the explanation of
   expectations. Simple reversal of a regression
   equation in these circumstances is not possible.
2. Given the earlier models, the dependent variable
   investigated was the change in the logarithms of
   the house price index data over the next three
   months, explained by relevant previous price
   changes and logistic survey variables over the
   period of 2000 M 10 to 2009 M10.
3. The list of variables with the Halifax series is
   followed by the alternative observations, the
   Nationwide, on the next two slides.
4. When comparing the two models, the Nationwide one
   seems slightly superior statistically, explaining
   ninety-four per cent of the variation.

12
Halifax Forecasting Model

• The dependent variable is the three months
  forward-looking growth rate using the Halifax
  price index
• Explained by the RICS survey diffusion variable,
  log(au/(1-au)) and the lagged values of it at 11,
  13, 22, and 24 months
• The forward-looking growth rate lagged 1, 3, 4,
  6, 7, 9, 10, 12 and 13 months.

13
Nationwide Forecasting Model

• The dependent variable is the three months
  forward-looking growth rate using the Nationwide
  price index
• Explained by the RICS survey diffusion variable
  log(au/(1-au)) lagged 10, 14, 20, 21, 22, 23 and
  24 months
• The dependent variable lagged 1, 3, 4, 6, 7, 9,
  11, 12, 17, 22 and 24 months.

14
Comparison with an Alternative Form

• The logistic forecasting model of the Nationwide
  was compared with the Pesaran/Thomas (PT)
  procedure for generating price expectations. The
  PT procedure involves analysis using the
  backward-looking survey adjusted ups with the
  inclusion of lagged residuals to explain
  backward-looking price growth. The coefficients
  from that analysis are then used in conjunction
  with forward-looking adjusted ups and residuals
  to generate the expected forward-looking change
  in house prices.
• The root mean squared forecast error (RMSFE) was
  used to compare the three months models. The
  equation with the logistic function led to lower
  RMSFE. Results of these tests for the Nationwide
  model are shown in the paper.
• Given the complexity of the dependent variable
  that developed from Hendry methodology with
  further restrictions imposed on the model
  compared with the previous slide, it was decided
  to experiment with a lag length beyond 3 months.
  When the lag length was put to 12 months ahead,
  the equation next was derived.

15
The Nationwide Forecasting model Twelve Months
Ahead

• The dependent variable is the log change of
  forward-looking growth rate using the Nationwide
  price index over the next twelve months
• Explained by the RICS survey diffusion variable
  log(au/(1-au)) lagged 1, 2, 6, 12, 13, 15 and 23
• The dependent variable lagged 1, 6,10, 11, 12, 22
  and 23 months
• One dummy variable for December 2007
• The statistical model suggests that the Survey
  data contains more than just three months of
  information, and could well contain a yearly
  sequence of events. The answering practices of
  the surveyors require investigation.

16
Forecasts based on the Yearly Model of the
Nationwide Price Index

• The study updated the dataset to May 2010, when
  the Nationwide index was 337.4603
• The previous equation was revised
• The analysis made a forecast for June, then
  updated the data, and forecast for July, and
  revised the data again. This process continued
  until the forecast for December
• The forecasts are as follows
• June - 334.2268,
• July 332.6814,
• August 326.6403,
• September 327.0887,
• October 309.9837,
• November 297.3425,
• December 292.5050.
• The model is forecasting a general decline in the
  price index with the June value representing a
  turning-point in the data set.

17
The Policy Implications of The Work

1. The work shows precisely how the RICS survey can
   be used to forecast the Halifax and Nationwide
   house price indices.
2. Expectations of future house prices are important
   for buyers (will prices go up after purchase?),
   sellers (could more be derived by selling
   later?), builders (will the house started fetch a
   profit on completion?) and should be important
   for the Government and the Bank of England (for
   example, what will the effect on consumer
   spending be?).
3. Expectations influence mortgage lenders. If
   expectations are for falling prices, they are
   reluctant to lend and tend to ration credit
   because of the greater probability of default. If
   expectations are for rising prices, lenders tend
   to make loans to a greater range of borrowers
   because of the rising value of collateral.

18
Governments Housing Policy

1. It would be useful to supplement this study with
   a co-integration VAR analysis incorporating the
   RICS survey data, to investigate the housing
   market fundamentals which affect house price
   expectations most, in both the short and long
   runs.
2. This would be a useful study because the
   Government can only realistically affect market
   expectations indirectly, through market
   fundamentals.
3. The first target of policy historically has
   tended to be housing finance interest rates and
   mortgage lender regulation.
4. The second target historically has tended to be
   housing supply, using local authorities and
   planning approval.

19
Conclusions/Summary

• The study has focused on the process of the
  formation of expectations of house prices
  underlying the RICS Survey
• The empirical investigation suggests that there
  is a diffusion process, captured by the logistic
  model. This is in line with models of bounded
  rationality, where decision-making is uncertain,
  self-fulfilling, complex and costly
• The majority of agents in the market follow the
  few, the alphas of the pack, namely the Chartered
  Surveyors.