Can interest rates really control house prices?

Effectiveness and Implications for Macroprudential

Policy

Song Shi

School of Economics and Finance, Massey University

Jyh-Bang Jou

Graduate Institute of National Development, National Taiwan University

David Tripe

School of Economics and Finance, Massey University

Version: June 2014

Acknowledgement: The authors would like to thank attendees at seminars at Massey

University, the Reserve Bank of New Zealand, the 2012 New Zealand Finance
Colloquium and the 2013 Conference of the New Zealand Association of Economists
for comments on earlier versions of this paper. Thanks are also due to McDermott
Miller for provision of consumer confidence index data and ASB Bank for housing
confidence data. The authors also thank Andrea Bennett, Faruk Balli and Hatice Ozer-
Balli for their assistance.
The authors also thank the editor and an anonymous referee for comments which have
allowed us to significantly improve the paper.


Corresponding author. Ph +64 6 356 9099 ext 84038, fax +64 6 3505660, e-mail
[email protected]. Address: School of Economics and Finance, Massey University, Private
Bag 11- 222, Palmerston North 4442, New Zealand
 Can interest rates really control house prices?
Effectiveness and Implications for Macroprudential
Policy

Abstract

This paper investigates how changes in the central bank policy and retail

mortgage rates affected real housing prices in New Zealand during the period 1999-

2009. We find that real interest rates are significantly and positively related to real

housing prices, indicating that increases in the policy rate may not be effective in

depressing real housing prices. By testing interest rates, we also find some evidence of

housing price bubbles. Our findings suggest that the central bank could have limited

housing price bubbles if it had started to intervene in the housing market prior to 2003.

Our results set international exemplars for using policy rates or macroprudential tools

to cool the housing market, where the extent of policy rate adjustments is limited by

internal or external economic factors

Keywords: Housing market, house price bubbles, monetary policy, mortgage lending

rates.

JEL Codes: E52, E58, G21, R21

2
1. Introduction

House price levels are one of the big economic issues of the 21st century. There

was strong growth in house prices from 2000 through to the onset of the global

financial crisis (GFC) in 2007/08. This was not just a United States phenomenon, with

strong house price growth and subsequent busts also observed in countries such as

Ireland and Spain. Rapid growth in house prices was also seen in New Zealand

(Bollard and Smith, 2006), where real interest rates were positively correlated with

real house price growth, in contrast with the expected negative relationship.

House prices may affect not only the economy as a whole, but also the banking

system. Houses often form an important part of household asset portfolios, with this

particularly the case in New Zealand where housing represented 74% of gross

household assets at the end of 2012. Rising house prices may create a wealth effect

which may encourage households to increase spending, resulting in inflationary

pressures and macroeconomic instability (Crowe et al., 2013).1 Lending on housing

accounted for 56% of New Zealand total credit as at the end of 2012, which would

mean that losses on housing lending could have a negative impact on bank solvency.

These sorts of effects have been observed internationally, in such countries as the

United States, Ireland and Spain in the aftermath of the GFC (Crowe et al., 2013), and

there is a fear that another round of house price rises could lead to a recurrence of

previous problems with bank solvency.

Previous research has pointed out that monetary policy can impact on housing

prices (Bernanke and Gertler, 1995; Mishkin, 2007; Shiller, 2006). The extent of any

change and the time it takes for the change to have an effect is a more problematic

1
Although their study looks at different aspects of the New Zealand housing market to those raised in
this paper, Grimes and Aitken (2010) also affirm the importance of house price dynamics.

3
issue, however. What is the impact of policy rates on real mortgage rates, and how do

these real mortgage rates impact on house prices? Are interest rates, as Crowe et al.

(2013) suggest, a blunt instrument which may engender undesirable outcomes in other

parts of the economy? These are among the important questions in debates occurring

in a number of countries as house prices are again increasing, as the effects of the

GFC abate, leading to discussion of the potential role for macroprudential policies.

The concern of regulators and policy-makers in this situation is that a house price

boom might be followed by a bust, undermining bank safety and soundness.

Consequently, if a boom can be prevented by appropriate intervention, we might also

be able to avoid subsequent damage to the banking sector.

The focus of this paper is New Zealand, which saw rapid growth in house prices

during the period 2001-2007, accompanied by a very substantial increase in total

lending on residential mortgages. At the same time, we saw continuing increases in

the Reserve Bank of New Zealand (RBNZ)’s Official Cash Rate (OCR), with general

pass-through to the interest rates paid by mortgage borrowers. Figure 1 shows the

relationship between the OCR, mortgage rates and real house price growth in

Auckland City (New Zealand’s largest) between 1999 and 2009.

<Insert Figure 1 >

The effect of the GFC on New Zealand house prices was relatively mild

(nominal price reductions of 10-15%, on average, after one year of the GFC), with

losses now recovered in most parts of the country.2 More significant price increases

are now occurring in some parts of the country, which has engendered concern by the

RBNZ that the country might be in for another housing price boom as in the mid-

2
See Hunt (2013) for more review of the relevant history.

4
2000s, setting the banks up for another bust, which might be aggravated by relatively

high levels of household indebtedness.

The RBNZ’s response to this has been to implement loan-to-value ratio

restrictions as a macroprudential policy tool, on the basis that it does not want to

increase interest rates because of the effect that might have on the value of the New

Zealand dollar and thus on the competitiveness of the export sector (New Zealand has

a long history of current account deficits in the balance of payments). There is also a

question as to how effective interest rates would be as a means of restraining sharply

rising house prices, or more generally, how strong the impact of interest rates is on

house prices. For example, Taylor (2007, 2009) argued that low interest rates

contributed to a boom and bust in the USA,3 while Glaeser et al. (2010) found that

decreases in interest rates could only explain 20% of the increases in real house prices.

In this article, using the present value model described by Campbell and Shiller

(1988a, b), we show that the impact of real interest rate changes on the real housing

price is positive. The effects of interest rates on housing prices can, however, be

further complicated by mortgage borrowers’ hedging activities in choosing between

fixed and floating rate loans, in anticipation of OCR movements, and we also

investigate this issue.

The New Zealand environment and the RBNZ’s policy actions make it a

particularly interesting market for studying the effectiveness of interest rate policy. In

the next section we look at monetary policy and interest rates in New Zealand in more

detail, and at how they interact with each other and with the housing market. Section 3

presents the theoretical framework and outlines the econometric tools used. Section 4

3
See also McDonald and Stokes (2013), who show a negative relationship between short term interest
rates in the United States (the Fed funds rate) and nominal house prices.

5
describes the data while Section 5 reports the empirical results. Section 6 outlines

policy recommendations and concludes.

2. New Zealand interest rates

The RBNZ’s monetary policy mandate is an inflation target, currently specified

as a range of 1 to 3%. The key policy rate is the OCR, which defines what banks earn

if they deposit funds overnight or pay if they borrow from the RBNZ. Any borrowings

are on a repo basis, with a spread of 50 basis points between the borrowing and

lending rates. The OCR thus sets a benchmark for interbank overnight borrowing and

the short end of the yield curve, with market rates for longer maturities being

impacted by standard factors that influence the yield curve.

The OCR is reviewed 8 times a year, although on one occasion (19 September

2001), it was changed at other than a scheduled review date. Following any OCR

review, the dates for which are scheduled a year or more in advance, there should be

no further change for another 6 to 7 weeks. Overnight market rates should thus remain

close to the OCR until the next scheduled review. The inflation targeting approach

means that the OCR should be higher when inflation is higher, and real interest rates

should be less variable than nominal interest rates. The highest nominal rate reached

for the OCR was 8.25% between 26 July 2007 and 24 July 2008, while its lowest

level has been 2.5% between 30 April 2009 and 10 June 2010, and again from 10

March 2011 until an increase on 13 March 2014.

Longer term money market rates also move in response to the OCR, although the

yield curve was negatively sloping between 2004 and 2009. This reflected

expectations that the RBNZ would ease short-term interest rates in response to easing

6
inflation pressures (an outcome that was delayed, at least in part, because of booming

house prices). Longer term rates fell significantly after the middle of 2008.

New Zealand borrowers have choices as to lending interest rates. They can

borrow at floating rates, which can be changed at relatively short notice (one month or

less), or they can fix the interest rate on their mortgage borrowing for a period

between six months and five years.4 Floating rate borrowers can switch to fixed rates

at any time, but fixed rate borrowers can only change their arrangements at the end of

the period for which they have taken a fixed rate, unless they pay an early repayment

penalty (generally calculated on the basis of interest rate differentials). One of the

consequences of the use of fixed rate loans is a delay between changes in market rates

and what borrowers actually pay, dampening the effect of monetary policy changes on

household spending capacity.

Actual rates charged to retail mortgage borrowers for different periods to

repricing generally track movements in the underlying money market rates. Tripe et al.

(2005) found that, since the adoption of the OCR approach in 1999, key lending rates

had become more responsive to changes in underlying wholesale rates, and the RBNZ

monetary policy could thus be described as having become more efficient. Liu et al.

(2008) found that the introduction of the OCR increased the pass-through to floating

rates, but not to fixed mortgage rates.

This has been confirmed in exploratory data analysis undertaken as part of this

research, where we find stronger correlations between short-term mortgage rates and

the OCR than for longer term rates. Further analysis using Granger causality tests

4
As in the USA mortgage market, mortgage loans in New Zealand can be for 25 or 30 years, but the
usual maximum term for a fixed interest rate is 5 years.

7
uncovered bi-directional relationships, 5 which might reflect the keen competition

between banks. Announced changes in the OCR often provide banks with the reason

to adjust their retail interest rates in response to changes in underlying funding costs.6

On the other hand, banks might change shorter fixed term mortgage lending rates

prior to an OCR announcement because they anticipate a policy change or have to

respond to the actions of other lenders changing their mortgage rates. The relationship

with longer term rates is likely to be an indication of the way long-term rates predict

future shorter-term rates. This suggests that the yield curve is acting as a relatively

reliable predictor of future rates. If the OCR did not respond to changes in other

interest rates, it is likely that we would regard the implementation of monetary policy

as somewhat erratic.

No prior research has looked at how either real New Zealand money market or

retail rates impact on house prices. Our paper contributes to the literature by

investigating this issue.

3. Estimation strategies and the empirical models

3.1 Present value model

This paper follows the present value model to investigate how the real rental rate

and the real interest rate affect the real housing price. Shiller (2006) argued that house

prices should be equal to the present discounted value of future rents. A linear present

value model with a constant discount rate is thus written as follows:

5
Results are available from the authors on request.
6
See Tripe et al. (2005) and Cottarelli and Kourelis (1994).

8
where Pt is the current asset price at time t, Dt is the dividend or cash flow at time t

and R is the constant expected discount rate. On the right-hand side of Equation (1),

the first term is called the fundamental value, and the second term the rational price

bubble. When n is sufficiently large, the second term will converge to zero. The well-

known Gordon growth model is accordingly set as follows:

where G is the constant growth rate of cash flows and is less than R. When G is zero,

Equation (2) becomes:

This formula is widely used in the valuation of income producing properties. The

term R is referred as the capitalisation rate or investment yield in real estate. The

model implies that house prices are positively related to rental rates, but negatively

related to the household’s discount rate.

The assumption of a constant expected discount rate R is analytically convenient,

but inconsistent with the variation in investors’ expected rates of return over time. It is

logical to infer that time-varying discount rates are closely linked to the retail

mortgage rates prevailing in the housing market. Other caveats for successfully

9
applying Equation (3) are that R must reflect future rental growth, real interest rates

and the housing premium.

In contrast with the literature, Equation (3) disregards two sets of factors. The

first is demand fundamentals such as employment and income (Campbell et al., 2009;

Wheaton and Nechayev, 2008), and net immigration (PricewaterhouseCoopers, 2009);

the second is credit market terms such as the loan-to-value ratio and approval rates

(Glaeser et al., 2010). While data on credit market terms exists in the USA, detailed

information on these is not generally published by banks in New Zealand. 7

Furthermore, the impacts of these terms on housing prices are inconclusive. While

Khandani et al. (2009) and Wheaton and Nechayev (2008) find that they play a

pivotal role, Glaeser et al. (2010) find that they are minor factors. We follow

Campbell et al. (2009) in adding other economic variables to Equation (3) to better

forecast real interest rates, rental growth and the housing premium.

However, a recent UK study by Tse et al. (2014) suggests that house prices could

be fractionally integrated, i.e. house prices are stationary processes with long memory.

This is important because if house prices are indeed fractionally integrated we need to

modify our estimation strategy. We test this possibility of fractional integration by

applying the Zivot-Andrews (1992) unit root tests and the ARFIMA (0,d,0) fractional

integration test to our New Zealand data set.8 Using the former test, we find that log

real house prices are non-stationary with potential structural breaks. In addition, using

the latter test with potential breaks, we find that first differencing of house prices is

not over-differenced. Finally, we also find that when both house prices and rents are

fractionally differenced, there is almost no difference between the two time series.

7
There was, however, some recognition of a potential easing of credit standards, as instanced by
comments in Reserve Bank of New Zealand (2007), pp. 25-26.
8
The results are available from the authors on request.

10
Thus we specify our estimation models in the first-order difference form. The results

of the ADF (Augmented Dickey-Fuller) unit root test are attached in Appendix 1.

Taking the first-order difference and logarithm of Equation (3), adding

macroeconomic conditions and location differences, we obtain:9

∆ , ∑ ∆ , ∑ ∆ , ∑ ∆ ,

∑ , , , ∑ (4)

where is a constant, i refers to different cities, t denotes the time period, l is the

number of lags, pi,t , di,t and , are log real prices, log real rents and real mortgage

rates, respectively, X is a vector of economic variables, , denotes the percentage

change of ratio of floating rate loans to overall mortgage loans, , denotes the

potential structural break as indicated from the Zivot-Andrews unit root tests, Sj

denotes the monthly seasonal dummy variables ( S j  1 for month j , and S j  0

otherwise), is the white noise, and Δ denotes the first difference or percentage

change.

The economic variables include the log consumer confidence index,

unemployment rate and real household lending. All variables are measured in the first

difference or percentage change. The consumer confidence index is used to control for

the impact of consumer confidence on housing market activities. Unemployment rate

and real household lending variables are measured on the percentage change basis.

The unemployment rate is used to control for labour market conditions, while the

9
Our specification contrasts with those of Wheaton and Nechayev (2008) and Igan and Loungani
(2012), neither of which includes any lagged price difference variable in Equation (4).

11
variable for real household lending is used to control for available credit in the

housing market.

The optimum lag number l is identified by the Hannan-Quinn (HQ) information

criterion, which is considered to be superior for sample size over 120 (see, for

example, Khim and Liew, 2004). Following the method suggested by Vahid and

Engle (1993) in choosing the order of the VAR system, we estimate different lengths

in levels and select the one with the smallest HQ criterion. The optimum lag length in

levels as indicated by the HQ criterion is two, and we thus take one lag for Equation

(4). A similar lag length selection is used in Leung et al. (2013) in their study of the

effect of international commodity prices on housing prices in New Zealand and

Australia. The structural break is taken at September 2007 across all models.10

The above estimation equation could potentially be improved by using a vector

error correction model (VECM) to study the long-run relationship of interest rates and

housing prices, but two problems arise. The first is that the OCR regime has only been

in place since April 1999, making long-run analysis problematic. The second is that

the VECM estimations of the long-run relationship between the variables are not

stable because estimated long-run relationships change dramatically for different

localities. 11 To overcome these two problems, we use pooled OLS to explore the

relationships between the variables. The Chow (1960) test indicates that the data are

poolable with a common intercept and the same slopes across different cities.12

10
The Zivot-Andrews unit root tests generally indicate that structural breaks occurred between
September 2007 and March 2008. Having regard to the narrow range of dates, we adopt a common
break for all markets of September 2007. This immediately followed the events of August 2007 which
have been commonly identified as the start of the GFC, and are also broadly consistent with the
turnaround in housing confidence data indicated by the ASB Bank survey. The results are robust for
using different possible structural breaks, such as March 2008.
11
The results of the VECM estimations are available from the authors on request.
12
The results of Chow (1960) and Wald tests are available from the authors on request.

12
3.2 Hedging effect of mortgage rate changes

We further explore mortgage choice and its impact on housing prices as we

believe that the choices between fixed and floating rate loans have important policy

implications. If more borrowers choose fixed rates, changes in floating mortgage rates

will have less direct impact on housing prices. Figure 2 shows the ratio of the value of

floating rate to overall mortgage loans over time. Since the RBNZ introduced the

OCR in 1999, the proportion of floating rate loans by value fell from 40% to 12.5% in

2007. However, with reductions in the OCR since 2008, floating rates have become

more attractive to borrowers. The value of floating rate loans relative to total loans

exceeded 25% by the end of 2009, and 50% by March 2011.

<Insert Figure 2>

There is, however, a problem in using the OLS to estimate Equation (4). Follain

(1990) notes that mortgage choice (floating vs. fixed) is an endogenous variable,

correlated with other factors such as expectations of future interest rate changes, so

that Rt and in Equation (4) might be correlated. To accommodate this, we use the

differential between the 5 years fixed and floating mortgage rates as an instrumental

variable (IV) in a two-stage least squares (2SLS) regression to estimate Equation (4).

Our models are thus specified as follows:

The first stage:

, ∑ ∆ , ∑ ∆ , ∑ ∆ ,

∑ , , ∑ (5)

The second stage:

13
 ∆ , ∑ ∆ , ∑ ∆ , ∑ ∆ ,

∑ , , , ∑ (5’)

where , is the estimated value of Ri,t in Equation (5).

The instrumental variable ( , ) is believed to be correlated with , but is

assumed to be uncorrelated with the error term , in Equation (4). This is likely to be

the case as differences between the 5 year fixed and floating rates will most affect

households’ mortgage choice, as measured by the percentage change in the ratio of

floating rate loans to overall mortgage loans (measured by value). A higher positive

interest rate differential between the 5 year fixed and floating rates (i.e. fixed rates are

higher than the floating rate) will push more people onto a floating rate and vice

versa.13 On the other hand, the interest rate differential between the 5 year fixed and

floating rates should have minimal influence on housing price changes (i.e.

households seldom buy or sell houses based on the interest rate differentials between

floating and fixed rate mortgages). Thus the interest rate differential would not be

expected to be correlated with the error term , .

3.3 Rational expectations and bubbles

We also test for bubbles in the housing market, as the existence of a bubble could

impact on the effectiveness of the policy rates. Under rational expectations (Lucas

and Sargent, 1981) households will use all past information up to time period t to

approximate house price growth at time t+1. Following this strategy, we first estimate

Equations (5) and (5’) using all past information up to time t, then use the estimated

coefficients and t+1 information to forecast the house price change pt1 . In this

13
The heteroskedasticity-robust t test in the reduced form regression shows that the ratio of floating
rate loan to overall mortgage value is indeed an endogenous variable.

14
process, rational expectations of house prices are developed from a mixture of current

and past fundamental information. Similar estimation strategies were used by Clayton

(1996) to derive a rational expectations model for housing price volatility, forecasting

rents and other market fundamental data based on the time series properties of the

data. He assumed that rents follow an AR(4) process and used the Box-Jenkins

technique to forecast other exogenous variables such as net immigration and the stock

of newly completed but unoccupied homes. His conclusions were thus sensitive to

these in-sample results. In this study, we follow a more general approach to forecast

the expected house price change pt1 , which can be written as:

E, ∆p , , , ∆p , ∑ , ∆d , ∑ λ , ∆m ,

∑ Ψ, X , φ, , B ∑ ϕ, S, , (6)

where E , ∆p , is the expected next period house price change for city i using

information of other variables at both time period t and t+1, “ ” denotes the

estimated values using 2SLS as specified in Equations (5) and (5’) and , is the

idiosyncratic effect associated with the individual city.

In Equation (6) the expected house price change not only depends on changes of

variables at both times t and t+1, but also on changes of parameter estimates

(coefficients) at t-1 and t. In other words, households’ discount rates are not only

linked to mortgage rate changes, but also to changes in coefficients of other variables.

The difference between the actual and expected price change is:

, ∆ , , ∆ , (7)

15
 We examine the distribution and time series properties of , to find whether

housing price bubbles exist. Our hypothesis is that if households are rational with

varying discount rates, , must be small and its distribution should be close to

normal. If a rational bubble exists, it will generate a set of small positive , over

time, followed by a large negative excess return at the time of the crash (e.g.,

Blanchard and Watson, 1982). The distribution of , for this type of bubble will

therefore be leptokurtic. One concern of this estimation strategy for bubbles is that

current price information may contain a “bubble” component, thus undermining

testing for a bubble in subsequent periods. Nevertheless, the trend of , will

provide indications for the existence of a bubble, to which the policy rate (or some

macroprudential policy tool) might be directed.

For comparison, we also estimate housing price misalignment assuming that the

relationship between households’ discount rates and other variables are constant. A

similar approach is used by Igan and Loungani (2012) in estimating global housing

cycles. They first modelled housing price changes in terms of changes in fundamental

variables in a base period, and then used the parameter estimates to forecast future

house prices. Later they use the gap between actual house prices and their predicted

values as an indication for a price “bubble.” However, this approach is of limited use

for several reasons. First, the estimation model must be complete. Second, the chosen

base period is arbitrary as prices must be at their fundamental levels, i.e., prices are

not overvalued or undervalued during the base period. Finally, the relationship among

variables for determining households’ discount rates are assumed to be the same in the

future, i.e., parameter estimates obtained in the base period will not change over time.

16
4. Data Description

This research utilised a rich data set of 528,601 freehold (fee simple) open

market transactions of detached or semi-detached houses for six cities in New Zealand

between 1994 and 2009. House price movements for the selected cities were

estimated directly, using the repeated sales method at monthly intervals, from

transaction data, which was unique and not publicly available. The transaction data

was supplied by Quotable Value (QV), the official database for all property

transactions in New Zealand. The six cities are Auckland, North Shore, Waitakere,

Manukau (all of which are now part of an expanded Auckland super-city), Wellington

and Christchurch, which are chosen because they accounted for more than 50% of

New Zealand housing stock and sales volume.

It is important to consider sample sizes when measuring local house price

movements using the Case-Shiller (1987) weighted repeated sale (WRS) method. As

this method uses only repeated sales for index construction, the index is more prone to

sample selection bias than other methods that use all transaction data. Previous work

indicates that frequently traded houses (sold more than twice within a period) are

more likely to be “starter” houses or houses for opportunistic buyers (Clapp and

Giaccotto, 1992; Haurin and Hendershott, 1991). Previous studies also indicate that

the repeat sales index is prone to a systematic downward revision due to lagged sales

(Clapham et al., 2006). To minimise these problems, we measure local house price

indices over an extended time period from 1994 to 2009. Table 1 shows the

distribution of house sales and numbers of dwellings, both of which indicate that we

have sufficient repeated sales, minimizing sample selection bias. Note that we use real

house prices, defined as nominal house prices adjusted by the CPI (Consumers Price

Index).

17
 <Insert Table 1>

QV also produces a house price index, but on a quarterly basis. The QV index is

based on the Sale Price Appraisal Ratio (SPAR) method, which takes the ratios of

current sale prices over their previous assessed values to construct an index.

Compared with the quarterly index, our monthly price index unsmooths price

movements and increases the number of observations in the time series. Monthly

analysis also allows us to make more effective use of New Zealand interest rate data.

As the repeat sales method is vulnerable to outliers (Meese and Wallace, 1997),

we use prior knowledge to eliminate multiple sales where the second sale price is less

than 0.7 or more than 2.5 times the first sale price. Moreover, since the QV data

includes building consent information for all except for Auckland City, we can

eliminate pair sales where quality has changed, thus minimizing the constant quality

problem faced by the repeat sales method.14 In total, we exclude 15% to 24% of initial

pair sales from estimation of the final index, depending on local housing markets. We

ended our data set in 2009 because it was the latest year for which we held a complete

sale data set, prior to the establishment of the Auckland super-city. Our repeated sales

price indices for the six cities are presented in Figure 3.

< Insert Figure 3>

We obtain monthly rental data for detached or semi-detached houses from the

Tenancy Services Division of the Ministry of Business, Innovation and Employment

14
Building consent data is collected for revaluation purposes only where QV is the valuation service
provider for the Council. This is not the case for Auckland City.

18
(MBIE)15 in New Zealand. Under the Residential Tenancies Act, all tenancy bonds

must be lodged with the MBIE within 23 working days from the start of the tenancy.

The bonds normally amount to two or three weeks of rent payable under the tenancy.

The DBH rental data is transaction based and very comprehensive in recording market

rents for all new residential tenancies.

For each local housing market, we use the monthly median rent, which is usually

for a 3-bedroom house. We use rental data for houses to proxy the user cost or

“imputed rent” of owning for the following reasons. First, we are unable to observe

the true user cost of owning a house. Even though we could estimate it (Hendershott

and Slemrod, 1983; Himmelberg et al., 2005), we would inevitably introduce

measurement errors. Second, the proportion of rental housing in the New Zealand

housing stock is large and increasing over time. By 2004 rental housing comprised

around 30% of the national housing stock.16 Thirdly, private sector rental houses and

owner-occupied houses tend to substitute for each other, and their prices do not differ

substantially. The survey by Hargreaves and Shi (2005) shows that on average rental

house prices lie between the open-market median and lower quartile house prices.17

Finally, we obtain the OCR, retail residential mortgage lending rates and values

of outstanding mortgage loans from the RBNZ. We use a monthly average OCR.

Retail interest rates include floating (or adjustable) rates, and rates fixed for 6 months,

1, 2, 3, 4 and 5 years. For the whole period of this study, fixed rate loans accounted

for the majority by value of all housing loans. We also use real interest rates, which

are defined as nominal interest rates adjusted by the CPI.

15
Previously the Department of Building and Housing.
16
Although New Zealand has traditionally had a high rate of home ownership, this rate declined
between 1996 and 2006. Analysis of census data from Statistics New Zealand shows that in 1996,
70.7% of households owned their dwellings, but it fell to 67.8% in 2001 and 66.8% in 2006.
17
Where the proportion of rental properties is high, rental houses are not restricted to less expensive
suburbs. In fact, rental housing has increased across all established suburbs across all New Zealand
cities.

19
 For the period from 1999 to 2009, for mortgage loans, we use data for household

lending from the RBNZ’s data table C5. Unemployment rate data comes from

Statistics New Zealand. Consumer confidence index data were obtained from

McDermott Miller, as part of a long-run data series that they provide for the Westpac

Bank Economics team. Summary statistics for the data are shown in Table 2.

<Insert Table 2>

5. Empirical Results

5.1 Housing prices and retail mortgage rates

Table 3 reports the relationship between house prices and mortgage rates based

on the 2SLS model specified in Equations (5) and (5’). The results generally support

our hypothesis that changes in the unemployment rate are negatively related to

changes in house prices, while changes in rents, consumer confidence and housing

lending are positively related to changes in house prices. 18 It also shows that the

current period house price growth is negatively correlated with the last period house

price changes, which indicates that we should include the lagged house price change

in the model.

However, our results show a strong positive relationship between interest rates

and house price growth for both floating and fixed rates (significant at the 1% level),

once household mortgage choices are controlled for.19 For example, a one percentage

18
The more money that goes into the housing market, the higher housing prices will be. On the other
hand, higher interest rates should dampen the demand for mortgage loans. However, we face the
problem of identifying causation, as it could be either that higher house prices cause increases in
housing lending or that increased housing lending leads to higher house prices.
19
The OLS regression results are included in Appendix 2. Compared to the 2SLS model used in this
study, it clearly shows the importance to include the household’s mortgage choice in the model. The

20
point increase in real floating rate (column 1 in Table 3) will result in 1.72% increase

in real house prices. The results indicate that households’ real discount rates have a

positive effect on real housing prices, which differs significantly from other housing

markets worldwide. Amongst all mortgage rates, 1 and 3 years real fixed-rates show a

relatively large impact on housing prices. However, the effects are not substantially

different between the floating and fixed rate mortgages, which could be due to the

relatively short fixed interest rate period up to a maximum of 5 years in New Zealand.

Our findings are in line with recent work on the USA market by Miles (2014), who

found the long term rate highly significant for housing while the short term fed funds

rate was not (although the relationship was negative, rather than positive as in our

case).

<Insert Table 3>

Our findings may reflect the shape of the yield curve, the mix of fixed and

floating rate lending, and the average size of fixed and floating rate loans. During the

potential bubble period (2004 to 2006), the yield curve was consistently negatively

sloping, with longer term fixed rates consistently cheaper than floating rates. This

meant that borrowers who wanted significant amounts of debt were often encouraged

to take on fixed-rate loans, both to reduce immediate debt servicing costs, and to

reduce their exposure to the risk of future interest rate increases. We thus saw an

increase in the relative proportion of fixed rate lending, while the size of the average

fixed rate loan (between $115,000 and $130,000 between 2004 and 2006) was much

larger than the size of the average floating rate loan (between $50,000 and $55,000

interest rates (mt), consumer confidence (CCIt), mortgage choice (Rt) and structural break (Bt) all
become significance at the 0.01 level in the 2SLS model.

21
over the same period). Borrowers with floating rate loans were likely to be less

concerned about the effects of interest rate changes as their smaller amount of debt

meant that the impact was going to be relatively small. Serious borrowing was

undertaken at fixed rates, and even though fixed rates increased through this period, it

was not enough to discourage borrowers, who perceived that property prices would

continue to increase.

5.2 Looking for bubbles

We also test whether the time series properties of ϵ , in Equation (7) exhibit a

pattern of bubbles. To estimate ϵ , , we first calculate the coefficients of Equation

(5’) using data from April 1999 to December 2002, and we then calculate ϵ , for

January 2003 based on the coefficients estimated up to December 2002. To calculate

ϵ, for February 2003, we re-estimate the coefficients of Equation (5’) using all

information from April 1999 to January 2003, and so on. We calculate the residual

term ϵ , for each of the six cities and seven mortgage rates, a total of 42 time series

of ϵ , from January 2003 to December 2009. The length of base period chosen for

the estimation is somewhat arbitrary. If it is too short, the first few periods’ estimated

ϵ, will not be reliable, whereas if it is too long, we will end up with fewer

observations for analysis. Table 4 provides results for those estimated ϵ , .

<Insert Table 4 >

Table 4 shows that the forecasting errors ϵ , for most cities and mortgage rates

are negative and close to zero. This is somewhat surprising given the rapid increase in

housing prices during the period studied. Results for kurtosis, confirming that

22
distributions, except for Manukau City, are generally leptokurtic, provide some

evidence for bubbles.20 A possible explanation is that the “bubble” might have started

earlier than 2003, and would therefore not be fully observed in the analysis. To

explore the price evaluating process, we break the summarised forecasting errors

ϵ, down into different time periods, to help us see how house prices evolve.

Table 5a shows the numbers of positive and negative forecasting errors in each

calendar year from 2003 to 2009, based on the floating mortgage rate. The results

show that house price appreciation slowed during 2004 to 2006, picked up in 2007

before the global financial crisis, slowed in 2008 and rapidly picked up in 2009.

Similar patterns are observed to varying degrees across all cities. The results are

consistent with the actual house price movements shown in Figure 3, which suggest

that if a bubble existed, it might have started prior to 2003.21 The findings provide

strong evidence that the best time for the policy rate to impact on the housing market

(pre-empting the bubble) was prior to 2003. From 2004 to 2006, the housing market

seemed to reach a new equilibrium at the high level. In 2007/2008, the market started

to adjust downward due to the global financial crisis, but it has recovered since 2009.

Table 5b shows the results based on the 5 year fixed mortgage rate, which largely

confirm the results in Table 5a.

<Insert Table 5a and 5b>

20
To show the reliability of our prediction, we plot those actual and estimated price changes over time
in Appendix 3. The results show that the predicted price changes are closely in line with the actual
price changes except for 2003 and during the global financial crisis in 2007/2008. For some cities, the
predicted price changes in 2003 may not be reliable due to the short base time period for forecasting.
21
By contrast, Hunt (2013) suggests that macroprudential tools might have been employed by the
RBNZ from around 2005 on.

23
 We also compare our approach to that suggested by Igan and Loungani (2012) to

estimate the price gap in levels. For comparison, we choose the same base period

from April 1999 to December 2002 to estimate Equations (5) and (5’). Using the

parameter estimates obtained in Equation (5’), we forecast future price changes from

January 2003 up to December 2009 for each city. Setting the price index in December

2002 at 100, we convert the price changes to price levels (indices). The forecast price

indices are then compared to the actual price indices to calculate price gaps.

Table 6 shows the estimated price gaps relative to the actual price indices, in

percentages based on different mortgage rates. For example, as indicated by the

floating rate (column 1), real house prices in North Shore City were about 25% higher

in 2005, 29% in 2006, 35% in 2007, 33% in 2008 and 34% in 2009. Overall, house

prices in Christchurch City appreciated the most, followed by Auckland City and

Wellington City. The gaps between the forecast and actual prices indicate moderate

price misalignments during 2005 and 2009. However, the results are subject to the

base period chosen for the comparison, with rapid house price appreciation having

occurred prior to 2003, leading us to believe the RBNZ should have intervened earlier.

<Insert Table 6>

6. Policy implications and conclusions

This paper has investigated how changes in the policy rate and retail mortgage

rates affect real housing prices in New Zealand during 1999-2009. We find that real

fixed interest rates do not have the expected negative effect on real housing prices,

after controlling for household mortgage choices and other economic conditions such

as the effect of real rental rates, unemployment rates, consumer confidence and

24
housing credit. Thus, increases in the policy rate did not depress real housing prices

during 1999-2009.

We also find that both the quantity of housing lending and the mix of fixed and

floating rate lending matter, with the latter also impacted by the slope of the yield

curve. This is not easy for governments or the RBNZ to control – intervention at the

long end of the yield curve would make the implementation of monetary policy a

much more complicated and expensive exercise. Against such a background, use of

macroprudential policies to mitigate some of the excesses of housing prices may be a

more attractive option.

Acceptance of findings that the connection between the policy rate (or short term

rate) and housing prices could be positive in a bubble environment has important

policy implications. Glaeser et al. (2010) argue that decreases in the real interest rate

could only explain 20% of increases in the real housing price, if one allows the

interest rate to be mean reverting. Dokko et al. (2011) also follow this argument.

Looking at 17 OECD countries, they find that decreases in policy rates in these

countries were not the main reason for housing price inflation during the mid-2000s.

Miles (2014) also finds that monetary policy rates do not necessarily have a strong

impact on house prices, while Crowe et al. (2013) also argue that monetary policy

tools are inclined to be ineffective at dealing with house prices, unless increases in

policy rates are particularly large.

Given that fixed mortgage rates have a larger impact on house prices than the

floating rate, and households can choose between floating and fixed mortgage rates,

we suggest that policy makers should consider macro-prudential tools such as

increasing down payment levels or capping loan-to-value ratios to influence the

housing market, particularly when the policy rate is directed at inflation targeting

25
while the national economy is exposed to global factors. This issue is particularly

important for New Zealand, where interest rate increases are constrained by other

national and global economic factors, such that an upward movement in the policy

rate, while other countries’ interest rates are relatively low, would adversely impact

the country’s balance of payments current account.

Our bubble tests under rational expectations show some but less severe incidence

of housing bubbles during the last decade, even though real interest rates were

positively correlated to real house price changes. This raises several questions. One

possibility is that the housing price “bubble” in New Zealand might have soft landed

during the GFC period in 2007/2008. High price levels after 2005 compared to the

base period prices from 1999 to 2002 may be simply suggesting that house price

levels moved from a low value regime to a high value regime, reflecting households’

changing discount rates. Labour income is taxed at 33% (the highest bracket), but

there is no housing capital gains tax in New Zealand. This might affect households’

expectation of return from owning. While a capital gains tax may be a tool to prevent

a bubble, it is difficult to implement in practice, as demonstrated by the failure of 8

tax working groups and inquiries (established by successive governments) between

1967 and 2010 to lead to any such tax being enacted (Huang and Elliffe, 2011). A key

further question for policy makers is whether they really want house prices to fall.

One of the limitations in this research is the relatively short period studied. This

means that we have not seen as much variation in economic conditions as would be

desirable to provide more robustness to our bubble test. Could there have been some

special factors which might have contributed to our potentially surprising results? The

other side of this is that there is scope for further research with the passage of time, so

that we can observe whether households’ future expectations for housing are rational

26
against a background of greater diversity in both external and internal economic

outcomes.

27
 References

Bernanke, B., Gertler, M., 1995. Inside the black box: The credit channel of monetary
policy transmission. Journal of Economic Perspectives 9, 27-48.

Blanchard, O., Watson, M., 1982. Bubbles, rational expectations and financial
markets, NBER Working Paper Series No.945. National Bureau of Economic
Research, Cambridge, MA.

Bollard, A., Smith, M., 2006. Major global developments in the new millenium.
Reserve Bank of New Zealand Bulletin 69, 26-39.

Campbell, S.D., Davis, M.A., Gallin, J., Martin, R.F., 2009. What moves housing
markets: A variance decomposition of the rent–price ratio. Journal of Urban
Economics 66, 90-102.

Case, K.E., Shiller, R.J., 1987. Prices of single-family homes since 1970: New
indexes for four cities. National Bureau of Economics Research, Cambridge,
MA.

Campbell, J. Y. & Shiller, R. J., 1988a. The dividend-price ratio and expectations of
future dividends and discount factors. Review of Financial Studies 1, 195-227.

Campbell, J. Y. & Shiller, R. J., 1988b. Stock prices, earnings, and expected
dividends. Journal of Finance 43, 661-676.

Chow, G.C., 1960. Tests of equality between sets of coefficients in two linear
regressions. Econometrica 28 (3), 591-605.

Clapham, E., Englund, P., Quigley, J.M., Redfearn, C.L. 2006. Revisiting the past and
settling the score: Index revision for house price derivatives. Real Estate
Economics 34, 275-302.

Clapp, J.M., Giaccotto, C., 1992. Repeated sales methodology for price trend
estimation: An evaluation of sample selectivity. Journal of Real Estate Finance
and Economics 5, 357-374.

Clayton, J., 1996. Rational expectations, market fundamentals and housing price
volatility, Real Estate Economics 24, 441-470.

Cottarelli, C., Kourelis, A., 1994. Financial structure, bank lending rates, and the
transmission mechanism of monetary policy. IMF Staff Papers 41, 587-623.

Crowe, C.; Dell’Ariccia, G.; Igan, D. & Rabanal, P., 2013. How to deal with real
estate booms: lessons from country experiences. Journal of Financial Stability
9, 300-319.

Dokko, J., Doyle, B.M., Kiley, M.T., Kim, J., Sherlund, S., Sim, J., Van Den Heuvel,
S., 2011. Monetary policy and the global housing bubble. Economic Policy 26,
237-287.

28
Follain, J.R., 1990. Mortgage choice. AREUEA Journal 18, 125-144.

Glaeser, E., Gottlieb, J., Gyourko, J., 2010. Can cheap credit explain the housing
boom?, NBER Working Paper No. 16230. National Bureau of Economics
Research, Cambridge, MA.

Grimes, A., Aitken, A., 2010. Housing supply, land costs and price adjustment. Real
Estate Economics 38 (2), 325-353.

Hargreaves, B., Shi, S., 2005. A total returns index for investor housing in New
Zealand. Pacific Rim Property Research Journal 11, 253-267.

Haurin, D.R., Hendershott, P.H., 1991. House price indexes: Issues and results.
AREUEA Journal 19, 259-269.

Hendershott, P.H., Slemrod, J., 1983. Taxes and the user cost of capital for owner-
occupied housing. AREUEA Journal 10, 375-393.

Himmelberg, C., Mayer, C., Sinai, T., 2005. Assessing high house prices: Bubbles,
fundamentals and misperceptions. National Bureau of Economic Research,
Cambridge, MA.

Huang, C.-C., Elliffe, C., 2011. Hard to argue with benefits of capital gains tax, New
Zealand Herald, Auckland, New Zealand.

Hunt, C., 2013. The last financial cycle and the case for macro-prudential intervention.
Reserve Bank of New Zealand Bulletin 76 (2), 3-16.

Igan, D., Loungani, P., 2012. Global housing cycles. IMF Working Paper No.
WP/12/17.

Khandani, A., Lo, A.W., Merton, R.C., 2009. Systemic risk and the refinancing
ratchet effect, NBER Working Paper No.15362. National Bureau of Economic
Research, Cambridge, MA.

Khim, V., Liew, S., 2004. Which lag selection criteria should we employ? Economics
Bulletin 3 (33), 1-9.

Leung, C.K.Y., Shi, S., Tang, E.C.H., 2013. Commodity house prices. Regional
Science and Urban Economics 43, 875-887.

Liu, M.H., Margaritis, D., Tourani-Rad, A., 2008. Monetary policy transparency and
pass-through of retail interest rates. Journal of Banking & Finance 32, 501-
511.

Lucas, R.E., Jr., Sargent, T., J., 1981. Rational expectations and econometric practice.
The University of Minnesota Press, Minneapolis.

McDonald, J.F., Stokes, H.H., 2013. Monetary policy and the housing bubble. Journal
of Real Estate Finance and Economics 46, 437-451.

29
Meese, R., Wallace, N., 1997. The construction of residential housing price indices: A
comparison of repeated sales, hedonic regression, and hybrid approaches.
Journal of Real Estate Finance and Economics 14, 51-74.

Miles, W., 2014. The housing bubble: How much blame does the fed really deserve?
Journal of Real Estate Research 36, 41-58.

Mishkin, F.S., 2007. Housing and the monetary transmisison mechanism, Working
Paper 13518. National Bureau of Economic Research, Cambridge, MA.

PricewaterhouseCoopers, 2009. New Zealand monetary policy and the residential

housing market - a scoping study working paper. Centre for Housing
Research, Aotearoa New Zealand, Wellington.

Reserve Bank of New Zealand, 2007 (May). Financial stability review. Reserve Bank
of New Zealand, Wellington.

Shiller, R.J., 2006. Long-term perspective on the current boom in home prices. The
Economist's Voice 3.

Taylor, J.B., 2007. Housing and monetary policy, NBER Working Paper No.13682.
National Bureau of Economic Research, Cambridge, MA.

Taylor, J.B., 2009. The financial crisis and the policy responses: An empirical
analysis of what went wrong, NBER Working Paper No.14631. National
Bureau of Economic Research, Cambridge, MA.

Tse, C.B., Rodgers, T., Niklewski, J., 2014. The 2007 financial crisis and the UK
residential housing market: did the relationship between interest rates and
house prices change? Economic Modelling 37, 518-530.

Tripe, D., McDermott, J., Petro, B., 2005. A test of the response to a monetary policy
regime change in New Zealand. Contemporary Studies in Economics and
Financial Analysis 86, 453-467.

Vahid, F., Engle, R.F., 1993. Common trends and common cycles. Journal of Applied
Econometrics 8 (4), 341-360.

Wheaton, W.C., Nechayev, G., 2008. The 1998-2005 housing "bubble" and the
current "correction": What's different this time? Journal of Real Estate
Research 30, 1-26.

Zivot, E., Andrews, D.W.K., 1992. Further evidence on the great crash, the oil-price
shock and the unit-root hypothesis. Journal of Business and Economic
Statistics 10, 251-270.

30
Table 1: Number of dwellings and sales in the major cities in New Zealand, Jan. 1994 – Dec. 2009
Number of North Waitakere Auckland Manukau Wellington Christchurch
Shore City City City City City City Total
Sales
Dwellings Dwellings Dwellings Dwellings Dwellings Dwellings Dwellings Sales
1 17,155 15,155 35,581 19,471 13,030 33,098 133,490 133,490
2 11,294 10,929 18,136 13,133 8,061 20,915 82,468 164,936
3 6,105 5,891 8,028 6,662 3,950 10,943 41,579 124,737
4 2,396 2,357 2,962 2,733 1,428 4,524 16,400 65,600
5 790 838 949 899 446 1,529 5,451 27,255
6 209 232 228 309 71 417 1,466 8,796
7 60 51 74 98 12 100 395 2,765
8 8 11 11 24 2 18 74 592
9 3 5 3 12 0 3 26 234
>=10 3 4 0 3 1 4 15 196
Total 38,023 35,473 65,972 43,344 27,001 71,551 281,364 528,601
Percentage* 54.88% 57.28% 46.07% 55.08% 51.74% 53.74% 52.56% 74.75%
The percentages* include multiple sales.

31
Table 2: Summary statistics of raw data
Variables Description Date Mean SD Max Min Obs Intervals
Price index
North Shore 1994 - 2009 1907 598 2998 1000 192 monthly
Waitakere 1994 - 2009 1983 611 3104 985 192 monthly
Auckland 1994 - 2009 2181 746 3522 1000 192 monthly
Manukau 1994 - 2009 1890 564 2975 1000 192 monthly
Wellington 1994 - 2009 1943 694 3221 1000 192 monthly
Christchurch 1994 - 2009 1660 569 2697 1000 192 monthly
Rent index
North Shore 1994 - 2009 1391 248 1875 958 192 monthly
Waitakere 1994 - 2009 1404 204 1750 1000 192 monthly
Auckland 1994 - 2009 1433 218 1840 1000 192 monthly
Manukau 1994 - 2009 1350 205 1773 1000 192 monthly
Wellington 1994 - 2009 1410 276 2130 957 192 monthly
Christchurch 1994 - 2009 1334 271 1889 1000 192 monthly
Interest (%)
Floating 1994 - 2009 8.73 1.53 11.50 6.20 192 monthly
6 months 1999 - 2009 7.45 1.21 9.93 5.52 129 monthly
1 year 1999 - 2009 7.54 1.10 9.90 5.64 129 monthly
2 years 1999 - 2009 7.71 0.86 9.63 5.94 129 monthly
3 years 1999 - 2009 7.88 0.72 9.61 6.13 129 monthly
4 years 1999 - 2009 8.01 0.65 9.56 6.42 129 monthly
5 years 1999 - 2009 8.06 0.62 9.50 6.52 129 monthly
OCR (%) 1999 - 2009 5.98 1.48 8.25 2.50 129 monthly
Consumer confidence index 1994 - 2009 115.19 9.86 130.90 81.70 64 quarterly
Unemployment rate (%) 1994 - 2009 2.37 0.82 5.10 1.00 64 quarterly
Household lending ($millions) 1998 - 2009 103,000 38,833 167,942 53,614 139 monthly
Ratio of floating rate loan to overall mortgage values 1998 - 2009 0.29 0.11 0.43 0.12 139 monthly
CPI index 1994 - 2009 904 95 1095 758 64 quarterly

32
Both price and rent indices start from 1000 from January 1994. Household lending is in million New Zealand dollars. The CPI index is set at 1000 in June 2006. The
consumer confidence index starts 100 from the second quarter of 2012.

33
Table 3: The results of the two-stage least squares pool regression for housing prices and retail mortgage
rates, April 1999 – December 2009
Floating 6 months 1 year 2 years 3 years 4 years 5 years
γ0 -0.012 *** -0.012 *** -0.011 *** -0.012 *** -0.010 *** -0.010 *** -0.010 ***
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Δp i,t (-1) -0.248 *** -0.263 *** -0.271 *** -0.277 *** -0.276 *** -0.274 *** -0.274 ***
(0.042) (0.041) (0.038) (0.038) (0.038) (0.038) (0.038)
Δd i,t 0.029 0.030 0.029 0.024 0.024 0.025 0.025
(0.025) (0.024) (0.023) (0.023) (0.023) (0.023) (0.023)
Δd i,t (-1) 0.051 ** 0.051 ** 0.048 ** 0.046 ** 0.048 ** 0.048 ** 0.048 **
(0.025) (0.024) (0.023) (0.023) (0.023) (0.023) (0.023)
Δmt 1.724 *** 1.489 *** 1.901 *** 1.661 *** 1.794 *** 1.728 *** 1.715 ***
(0.542) (0.446) (0.434) (0.382) (0.385) (0.395) (0.386)
Δmt (-1) 0.005 0.035 -0.298 -0.017 -0.035 0.076 0.130
(0.348) (0.301) (0.288) (0.242) (0.248) (0.249) (0.254)
Et 0.002 -0.001 -0.001 -0.001 -0.001 0.000 0.000
(0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004)
Et (-1) -0.011 ** -0.008 ** -0.011 *** -0.010 *** -0.010 *** -0.010 *** -0.009 **
(0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004)
Ct 1.624 *** 1.629 *** 1.286 *** 1.286 *** 1.218 *** 1.330 *** 1.361 ***
(0.406) (0.384) (0.362) (0.357) (0.354) (0.355) (0.358)
Ct (-1) 0.412 0.309 0.445 0.417 0.493 0.422 0.412
(0.385) (0.369) (0.355) (0.352) (0.351) (0.350) (0.352)
ΔCCIt 0.061 * 0.067 * 0.061 * 0.076 ** 0.075 ** 0.069 ** 0.065 **
(0.036) (0.035) (0.033) (0.033) (0.033) (0.033) (0.033)
ΔCCIt (-1) 0.027 0.024 0.028 0.005 0.001 0.007 0.012
(0.037) (0.036) (0.034) (0.035) (0.035) (0.035) (0.035)
Rt 0.377 *** 0.342 *** 0.269 *** 0.258 *** 0.253 *** 0.265 *** 0.270 ***
(0.095) (0.076) (0.063) (0.063) (0.059) (0.061) (0.063)
Bt -0.010 *** -0.010 *** -0.009 *** -0.009 *** -0.010 *** -0.010 *** -0.011 ***
(0.003) (0.003) (0.002) (0.002) (0.002) (0.002) (0.003)
Seasonal controlled yes yes yes yes yes yes yes
Observations 768 762 762 762 762 762 762
The dependant variable is the log difference of real house prices. The results are estimated using the two stage-least squares
method in a pool regression of 6 cities. The instrumental variable (Ii,t) is the interest rate differential between the 5 years fixed rate
and floating rate. Results are reported for 7 mortgage rates respectively using the following models:
The first stage:

, ∑ ∆ , ∑ ∆ , ∑ ∆ , ∑ , , ∑

(5)

The second stage:

∆ , ∑ ∆ , ∑ ∆ , ∑ ∆ , ∑ , , , ∑

(5’)

where Ri,t refers to the percentage change of the ratio of floating rate loans to overall value of mortgage loans,
, is the estimated value of , in Equation 5 , γ0 is constant, i refers to different cities, t denotes the time period and Δ
denotes the first order difference or percentage change. pi,t, di,t and mi,t are log real prices, log real rents and real mortgage rates,
respectively, Xi,t represents the list of economic variables including the percentage change of unemployment rate (Et), the

the monthly seasonal dummy variables, and i,t is white noise. The results are robust to period heteroskedasticity and serial
percentage change of real house lending (Ct) and consumer confidence index (CCIt). Bi,t refers to the structural break, Sj denotes

correlation (Period SUR) test. Standard errors in parenthesis; Statistical significance: *<0.10, **<0.05, ***<0.01

34
Table 4: Summarised statistics of forecast errors for monthly house price changes, 2002 - 2009

Floating 6 months 1 year 2 years 3 years 4 years 5 years

North Shore City
Mean -0.003 -0.003 -0.002 -0.001 -0.002 -0.002 -0.002
Median -0.006 -0.004 -0.004 -0.004 -0.004 -0.004 -0.004
Maximum 0.061 0.061 0.064 0.064 0.068 0.066 0.067
Minimum -0.094 -0.075 -0.057 -0.041 -0.040 -0.041 -0.042
Std. Dev. 0.025 0.023 0.021 0.020 0.020 0.020 0.020
Skewness -0.489 -0.004 0.391 0.833 0.805 0.778 0.838
Kurtosis 5.173 4.218 3.856 4.274 4.495 4.185 4.485
Observations 84 84 84 84 84 84 84
Waitakere City
Mean -0.004 -0.001 0.000 0.000 0.000 0.000 0.000
Median -0.004 -0.001 -0.001 -0.001 -0.001 -0.001 -0.002
Maximum 0.044 0.048 0.046 0.046 0.043 0.044 0.045
Minimum -0.087 -0.049 -0.050 -0.044 -0.046 -0.046 -0.047
Std. Dev. 0.022 0.017 0.016 0.017 0.017 0.017 0.017
Skewness -0.940 0.151 0.077 0.136 0.229 0.207 0.262
Kurtosis 5.036 3.149 3.465 3.250 3.249 3.179 3.376
Observations 84 84 84 84 84 84 84
Auckland City
Mean -0.004 -0.002 -0.003 -0.002 -0.001 -0.001 -0.001
Median -0.002 0.000 -0.002 -0.001 -0.001 -0.001 -0.001
Maximum 0.048 0.046 0.057 0.060 0.053 0.051 0.051
Minimum -0.100 -0.068 -0.077 -0.068 -0.074 -0.071 -0.073
Std. Dev. 0.025 0.022 0.025 0.021 0.020 0.020 0.020
Skewness -0.816 -0.294 -0.329 -0.075 -0.277 -0.278 -0.342
Kurtosis 4.888 3.254 3.619 3.409 3.876 3.792 4.047
Observations 84 84 84 84 84 84 84
Manukau City
Mean 0.000 0.001 0.000 0.001 0.002 0.001 0.001
Median -0.001 0.003 -0.001 0.001 0.002 0.001 0.002
Maximum 0.060 0.050 0.050 0.048 0.046 0.046 0.047
Minimum -0.041 -0.046 -0.054 -0.042 -0.041 -0.040 -0.041
Std. Dev. 0.020 0.021 0.021 0.019 0.018 0.018 0.018
Skewness 0.201 -0.018 -0.083 0.078 0.073 0.094 0.104
Kurtosis 3.008 2.445 2.836 2.759 2.782 2.734 2.817
Observations 84 84 84 84 84 84 84
Wellington City
Mean 0.000 0.000 -0.002 -0.002 -0.002 -0.002 -0.002
Median -0.002 -0.003 -0.002 -0.003 -0.002 -0.002 -0.003
Maximum 0.075 0.081 0.078 0.078 0.080 0.081 0.081
Minimum -0.046 -0.044 -0.049 -0.051 -0.049 -0.049 -0.050
Std. Dev. 0.023 0.023 0.023 0.023 0.023 0.023 0.023
Skewness 0.391 0.536 0.452 0.482 0.550 0.532 0.532
Kurtosis 3.458 3.727 3.838 3.989 4.137 4.085 4.095
Observations 84 84 84 84 84 84 84
Christchurch City
Mean -0.005 -0.001 -0.002 -0.001 -0.001 -0.001 -0.001
Median -0.002 0.000 0.000 0.001 0.001 0.000 -0.001
Maximum 0.087 0.046 0.048 0.047 0.046 0.049 0.049

35
 Minimum -0.137 -0.070 -0.087 -0.048 -0.035 -0.033 -0.036
Std. Dev. 0.033 0.022 0.024 0.020 0.019 0.019 0.019
Skewness -1.482 -0.237 -0.495 0.117 0.289 0.374 0.341
Kurtosis 8.207 3.346 4.106 2.679 2.553 2.677 2.679
Observations 84 84 84 84 84 84 84
Forecast errors , are estimated using the following equation: , ∆ , , ∆ , where ∆ , is the actual
monthly house price change for the ith city at time t+1 and , ∆ , refers to the forecasted monthly house price change for
the ith city for time t+1. To calculate the forecasting value of , ∆ , , information up to the time t is used in the following
two-stage least squares model stated below Table 3. The estimation is rolling over the studied period, which lasted from January
2003 to December 2009.

36
Table 5a: Number of positive and negative forecast errors - floating rate
2003 2004 2005 2006 2007 2008 2009 overall
North Shore City
+ 5 2 2 3 7 6 10 35
- 7 10 10 9 5 6 2 49
Waitakere City
+ 5 2 4 4 7 5 10 37
- 7 10 8 8 5 7 2 47
Auckland City
+ 3 1 6 5 8 5 10 38
- 9 11 6 7 4 7 2 46
Manukau City
+ 3 2 7 4 6 8 9 39
- 9 10 5 8 6 4 3 45
Wellington City
+ 6 4 7 5 2 4 9 37
- 6 8 5 7 10 8 3 47
Christchurch City
+ 5 3 4 4 5 7 9 37
- 7 9 8 8 7 5 3 47
The table presents the number of positive and negative forecast errors based on the floating mortgage rate for each year from 2003
to 2009. The forecast errors are obtained from the results of Table 5 for the floating rate.

Table 5b: Number of positive and negative forecast errors – 5 year fixed rate
2003 2004 2005 2006 2007 2008 2009 overall
North Shore City
+ 6 1 2 3 7 5 9 33
- 6 11 10 9 5 7 3 51
Waitakere City
+ 4 0 7 3 8 5 11 38
- 8 12 5 9 4 7 1 46
Auckland City
+ 5 1 5 5 9 4 10 39
- 7 11 7 7 3 8 2 45
Manukau City
+ 4 5 7 4 8 7 9 44
- 8 7 5 8 4 5 3 40
Wellington City
+ 4 5 8 5 2 5 9 38
- 8 7 4 7 10 7 3 46
Christchurch City
+ 4 3 5 4 7 8 10 41
- 8 9 7 8 5 4 2 43
The table presents the number of positive and negative forecast errors based on the 5 year fixed mortgage rate for each year from
2003 to 2009. The forecast errors are obtained from the results of Table 4 for the 5 year fixed rate.

37
Table 6: Estimated price gaps
Floating 6 months 1 year 2 years 3 years 4 years 5 years
North Shore City
2005 25% 25% 25% 25% 25% 25% 24%
2006 29% 29% 29% 28% 28% 28% 28%
2007 35% 35% 35% 35% 35% 34% 35%
2008 33% 32% 31% 30% 31% 31% 31%
2009 34% 33% 32% 31% 32% 32% 33%
Waitakere City
2005 26% 27% 28% 28% 27% 27% 27%
2006 31% 31% 31% 31% 31% 31% 31%
2007 37% 37% 37% 37% 37% 37% 37%
2008 38% 34% 32% 33% 33% 33% 33%
2009 38% 33% 31% 31% 32% 32% 32%
Auckland City
2005 21% 21% 21% 21% 21% 21% 21%
2006 26% 25% 25% 25% 25% 25% 25%
2007 31% 31% 31% 31% 31% 31% 31%
2008 31% 27% 29% 27% 26% 25% 25%
2009 34% 28% 31% 28% 27% 27% 27%
Manukau City
2005 21% 22% 21% 20% 21% 21% 21%
2006 26% 28% 27% 26% 26% 26% 26%
2007 33% 34% 34% 33% 33% 33% 33%
2008 32% 31% 32% 29% 28% 28% 28%
2009 31% 29% 31% 28% 27% 27% 27%
Wellington City
2005 20% 20% 20% 20% 20% 20% 20%
2006 26% 26% 27% 26% 26% 26% 26%
2007 33% 33% 33% 33% 33% 33% 33%
2008 30% 30% 32% 31% 31% 31% 31%
2009 32% 32% 34% 33% 33% 33% 33%
Christchurch City
2005 40% 35% 35% 35% 35% 35% 35%
2006 47% 40% 40% 39% 39% 39% 39%
2007 51% 44% 44% 44% 43% 43% 43%
2008 56% 43% 43% 41% 41% 41% 41%
2009 64% 44% 44% 41% 41% 41% 41%
The table presents the estimated price gaps between the actual house prices and predicted house prices in percentage terms with
respect to different mortgage rates, based on the house price levels during 1999 and 2002. The estimation procedure is similar to
that in Igan and Loungani (2012). Using the parameter estimates obtained in the 2SLS during 1999 and 2002, we forecast future
price changes from January 2003 up to December 2009. Setting the price index in December 2002 at 100, we convert the price
changes to price levels (indices). The forecast price indices are then compared to the actual price indices to calculate price gaps in
percentage terms.

38
 Figure 1: Prices and interest rates – Auckland City, April 1999 – December 2009

Figure 2: OCR movements and the ratio of the value of floating loan to overall
mortgage loan, Apr. 1999 – Dec. 2009
10% .44

9% .40

8% .36

7% .32

6% .28

5% .24

4% .20

3% .16

2% .12
99 00 01 02 03 04 05 06 07 08 09

OCR
Ratio of the value of floating loan to overall mortgage loan

39
Figure 3: Log real house price indices measured by the weighted repeated sales method
North Shore City Waitakere City Auckland City
8.0 8.2 8.2

7.9
8.0 8.0
7.8

7.7 7.8 7.8

7.6
7.6 7.6
7.5

7.4 7.4 7.4

7.3
7.2 7.2
7.2

7.1 7.0 7.0

94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09

Manukau City Wellington City Christchurch City

8.0 8.2 7.9

7.9 7.8
8.0
7.8
7.7
7.7 7.8
7.6
7.6
7.6 7.5
7.5
7.4
7.4 7.4
7.3
7.3
7.2
7.2 7.2

7.1 7.0 7.1

94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09

40
Appendix 1: The ADF unit root tests
Variables Level Level 1st Difference 1st Difference 1st Difference
(constant) (constant & trend) (constant) (constant) (constant)
HQ criterion HQ criterion HQ criterion SIC criterion AIC criterion
Log real house prices
North Shore -1.123 -2.056 -2.956 ** -2.956 ** -2.956 **
Waitakere -0.894 -2.044 -3.077 ** -3.967 *** -3.077 **
Auckland -1.217 -1.248 -4.213 *** -4.213 *** -3.466 ***
Manukau -1.015 -2.170 -2.924 ** -12.587 *** -2.924 **
Wellington -0.705 -2.068 -3.191 ** -14.278 *** -3.191 **
Christchurch -0.924 -1.979 -2.696 * -3.509 *** -2.696 *
Log real rents
North Shore -0.873 -1.463 -14.857 *** -14.857 *** -4.908 ***
Waitakere -1.157 -1.771 -21.771 *** -21.771 *** -21.771 ***
Auckland -1.989 -2.427 -20.040 *** -20.040 *** -20.040 ***
Manukau -1.007 -3.745 ** -10.291 *** -10.291 *** -10.291 ***
Wellington 0.797 -5.654 *** -3.582 *** -11.817 *** -3.582 ***
Christchurch -1.347 -2.405 -2.003 -11.822 *** -2.003
Real mortgage rates
Floating -2.270 -1.958 -4.601 *** -4.601 *** -4.775 ***
6 months -2.303 -2.094 -5.166 *** -7.793 *** -5.166 ***
1 year -2.461 -2.360 -5.377 *** -7.510 *** -5.377 ***
2 years -2.501 -2.458 -8.821 *** -8.821 *** -6.331 ***
3 years -2.763 * -2.765 -9.208 *** -9.208 *** -6.545 ***
4 years -2.798 * -2.831 -9.314 *** -9.314 *** -6.873 ***
5 years -2.749 -2.759 -9.632 *** -9.632 *** -9.632 ***
Real OCR -1.943 -1.677 -4.345 *** -4.345 *** -4.345 ***
Log consumer confidence index -2.575 -2.723 -3.232 ** -4.290 *** -3.232 **
Unemployment rate -4.411 *** -4.766 *** -3.646 *** -3.646 *** -3.646 ***
Real household lending -0.372 -2.248 -2.691 * -2.691 * -2.691 *

41
 Ratio of floating rate mortgage loans -1.038 -0.051 -4.417 *** -4.417 *** -4.417 ***
The maximum lag length is set at 12. For the unemployment rate, the results are based on the percentage change not the 1st difference of unemployment rate. Similar
percentage measurements are also taken to the real household lending and ratio of floating rate mortgage loans variables. All other variables are taken at 1st difference in this
study. Sample is between 1999m4 and 2009m12.

42
Appendix 2: The OLS regression results
Floating 6 months 1 year 2 years 3 years 4 years 5 years
γ0 -0.010 *** -0.010 *** -0.010 *** -0.010 *** -0.009 *** -0.009 *** -0.009 ***
(0.002) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Δpi,t (-1) -0.263 *** -0.262 *** -0.267 *** -0.269 *** -0.268 *** -0.266 *** -0.267 ***
(0.035) (0.036) (0.036) (0.036) (0.036) (0.036) (0.036)
Δdi,t 0.035 * 0.037 * 0.035 * 0.033 0.033 0.034 0.034
(0.021) (0.021) (0.021) (0.021) (0.021) (0.021) (0.021)
Δdi,t (-1) 0.058 *** 0.059 *** 0.056 *** 0.054 ** 0.055 *** 0.056 *** 0.056 ***
(0.021) (0.021) (0.021) (0.021) (0.021) (0.021) (0.021)
Δmt -0.026 0.044 0.678 ** 0.604 ** 0.689 ** 0.514 * 0.567 **
(0.315) (0.294) (0.312) (0.268) (0.273) (0.273) (0.266)
Δmt (-1) -0.602 ** -0.258 -0.323 -0.025 -0.146 -0.074 -0.079
(0.271) (0.260) (0.268) (0.227) (0.233) (0.231) (0.233)
Et -0.005 -0.004 -0.004 -0.004 -0.004 -0.004 -0.004
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Et (-1) -0.007 ** -0.007 ** -0.010 *** -0.010 *** -0.009 *** -0.009 *** -0.009 ***
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Ct 1.168 *** 1.253 *** 1.103 *** 1.140 *** 1.100 *** 1.172 *** 1.157 ***
(0.333) (0.331) (0.334) (0.333) (0.333) (0.331) (0.332)
Ct (-1) 0.207 0.149 0.246 0.186 0.254 0.189 0.205
(0.324) (0.323) (0.327) (0.326) (0.327) (0.325) (0.326)
ΔCCIt 0.041 0.042 0.042 0.049 0.050 0.047 0.046
(0.030) (0.031) (0.031) (0.031) (0.031) (0.031) (0.031)
ΔCCIt(-1) 0.056 * 0.057 * 0.053 * 0.043 0.040 0.046 0.046
(0.031) (0.031) (0.031) (0.032) (0.032) (0.032) (0.032)
Rt -0.025 -0.008 0.012 0.017 0.021 0.016 0.017
(0.023) (0.024) (0.024) (0.024) (0.025) (0.025) (0.025)
Bt -0.002 -0.002 -0.002 -0.003 -0.003 -0.003 * -0.003 *
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Seasonal controlled Yes Yes Yes Yes Yes Yes Yes
Observations 774 762 762 762 762 762 762
R 0.229 0.226 0.230 0.230 0.231 0.229 0.230

43
The dependant variable is log difference of real house price. The results are estimated using the OLS model in a pool regression of 6 cities as a comparison to the 2sls model. The OLS model is
defined as follows: ∆ , ∑ ∆ , ∑ ∆ , ∑ ∆ , ∑ , , , ∑ where γ0 is constant, i refers to different cities, t
denotes the time period and Δ denotes the first order difference or percentage change. pi,t, di,t and mi,t are log real prices, log real rents and real mortgage rates, respectively, Xi,t represents the

percentage change of the ratio of floating rate loans to overall value of mortgage loans, Bi,t refers to the structural break, Sj denotes the monthly seasonal dummy variables, and i,t is white
list of economic variables including the percentage change of unemployment rate (Et), the percentage change of real house lending (Ct) and consumer confidence index (CCIt). Ri,t is the

noise. The results are robust to period heteroskedasticity and serial correlation (Period SUR) test. R denotes the adjusted R-squared. The sample period is from April 1999 to December 2009.

44
Appendix 3: Actual house price changes and predicted house price changes, 2003m1 to 2009m12

Panel A: Floating rate

North Shore Waitakere Auckland

.20 .20 .20

.15 .15 .15

.10 .10 .10

.05 .05 .05

.00 .00 .00

-.05 -.05 -.05

-.10 -.10 -.10

I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009

Actual house price changes Actural house price changes Actual house price changes
Predicted house price changes Predicted house price changes Predicted house price changes

Manukau Wellington Christchurch

.20 .20 .20

.15 .15 .15

.10 .10 .10

.05 .05 .05

.00 .00 .00

-.05 -.05 -.05

-.10 -.10 -.10

I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009

Actural house price changes Actural house price changes Actural house price changes
Predicted house price changes Predicted house price changes Predicted house price changes

45
Panel B: 5 years fixed rate
North Shore Waitakere Auckland

.06 .06 .06

.04 .04 .04

.02 .02 .02

.00 .00 .00

-.02 -.02 -.02

-.04 -.04 -.04

-.06 -.06 -.06

I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009

Actural house price changes Actural house price changes Actural house price changes
Predicted house price changes Predicted house price changes Predicted house price changes

Manukau Wellington Christchurch

.06 .06 .06

.04 .04 .04

.02 .02 .02

.00 .00 .00

-.02 -.02 -.02

-.04 -.04 -.04

-.06 -.06 -.06

I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009 2003 2004 2005 2006 2007 2008 2009

Actural house price changes Actural house price changes Actural house price changes
Predicted house price changes Predicted house price changes Predicted house price changes

46