-Master thesis: International Economics &
Business-Monetary Policy and Asset Prices
The influence of the short-term interest rate on property prices
Anne M. Nijenhuis S1284282
University of Groningen Faculty of Economics and Business
Groningen
1stSupervisor: Dr. D.J. Bezemer 2ndSupervisor: Dr. G.J. Lanjouw
Keywords: Property Price, Short-Term Interest Rate, Long-Term Interest Rate, Asset Price,
Abstract
TABLE OF CONTENTS
1. INTRODUCTION 6
1.1 MAIN RESEARCH QUESTION 8
2. LITERATURE REVIEW 10
2.1 THE ROLE OF ASSET PRICES IN MONETARY POLICY 10
2.2 ARGUMENTS IN FAVOR AND AGAINST RESPONDING TO AN ASSET PRICE BUBBLE. 14
2.3 MONETARY POLICY REACTION FUNCTIONS 16
2.4 EMPIRICAL RESEARCH 19
2.5 SHORT-TERM INTEREST RATES AND ASSET PRICES 21
2.5.1 EQUITYPRICES 22
2.5.2 PROPERTY PRICES 22
3. EMPIRICAL RESEARCH 25
3.1 ECONOMIC MODEL 25
3.2 METHODOLOGY USED FOR RESEARCH 27
3.3 THELEASTSQUARESDUMMYVARIABLEMETHOD ORFIXEDEFFECTS MODEL 28
4. TESTING THE ASSUMPTIONS 29
4.1. DESCRIPTIVE STATISTICS AND NORMAL DISTRIBUTION 29
4.2 COLLINEARITY ANDMULTICOLLINEARITY 29
4.3 STATIONARITY 29 4.4 HETEROSKEDASTICITY 30 4.5 AUTOCORRELATION 31 4.6 REVERSECAUSALITY 31 5. RESULTS 32 5.1 RESULTSSAMPLE1971Q3-2004Q3 32 5.2 RESULTSSAMPLE1985Q1-2004Q3 33 6. CONCLUSION 35
6.1 LIMITATIONS ANDFUTURERESEARCH 36
TABLEA1: VARIABLE DEFINITION AND SOURCES 38
TABLEA2: SAMPLESPECIFICATION 39
TABLEA3: AGGREGATE ASSET PRICE BOOMS IN SELECTEDOECDCOUNTRIES(1970-2004) 39
TABLEA4: MORTGAGE MARKET DEREGULATION MEASURES 40
TABLEA5: COUNTRIES AND THEIR BANKING CRISES 41
B1 DESCRIPTIVESTATISTICS WITH OUTLIERS(SAMPLE1971Q3-2004Q3) 42 B2 DESCRIPTIVESTATISTICS WITHOUT FAR OUTLIERS(SAMPLE1971Q3-2004Q3) 42 B3 DESCRIPTIVESTATISTICS WITH OUTLIERS(SAMPLE1985Q1-2004Q3) 43 B4 DESCRIPTIVESTATISTICS WITHOUT FAR OUTLIERS(SAMPLE1985Q1-2004Q3) 43
B5 CORRELATIONMATRIX(SAMPLE1971Q3-2004Q3) 44
B7 VARIANCEINFLATIONFACTOR(SAMPLE1971Q3-2004Q3) 46 B8 VARIANCEINFLATIONFACTOR(SAMPLE1985Q1-2004Q3) 46
B9 TEST FOR UNIT ROOT IN PANEL DATA 47
C1 REGRESSION RESULTS(SAMPLE1971Q3-2004Q3) 48
C2 REGRESSION RESULTS(SAMPLE1985Q1-2004Q3) 49
C3 REGRESSION RESULT WITHOUT INTERACTION TERM(SAMPLE1985Q1-2004Q3) 50
1. Introduction
Today, the US economy suffers from the subprime crisis; a mortgage crisis, affecting not only people, but also companies.
After 9-11, the day the twin towers were attacked in the United States, the American Central Banks kept interest rates low, while the US economy recovered from its recession. ‘It was our job to unfreeze the American banking system if we wanted the economy to function. This required that we keep rates modestly low,’ thus Greenspan1. As a result, housing activity in the US was booming, but when interest rates increased again, demand fell back.
Speculations on whether this crisis could have been prevented and what the Federal Reserve (Fed), United States’ central bank, should do next can be read on a daily, or even hourly basis in American newspapers. Some people blame the Fed on its lack of policy action when the property prices increased sky-high. Others point their finger at Greenspan, the Fed’s former chairman, who encouraged low-interest loans in 2002, so people could buy homes and the US could climb out of its recession after 9-11.
The situation as it now occurs in the United States is not a new phenomenon. In history various asset price booms have occurred. Amongst others, countries in Scandinavia, Japan, the Netherlands and the United Kingdom, suffered from boom-and-bust cycles (Bernanke and Gertler, 1999). A very famous asset price bubble is the one in Japan. From the latter half of the 1980s till the beginning of the 1990s, Japan experienced an asset price bubble. During this time, there was a rapid rise in asset prices, money supply and credit increased sizably and economic activity expanded dramatically (Okina, Shirakawa and Shiratsuka, 2001).
In the beginning of the 90s, the central bank of Japan decided to increase the discount rate and to control commercial bank lending by imposing aggregate limits to supplied credits. In addition a ‘capital adequacy requirement’ was introduced, which caused the Japanese bank loan supply to shrink (Horská, 2002). Consequently, the bubble collapsed; asset prices plummeted, huge nonperforming assets and resulting difficulties faced by financial institutions were accumulated and a recession started (Okina, Shirakawa and Shiratsuka, 2001).
1
But how are asset prices different from prices of current goods and services? Plosser (2007) describes asset prices as forward looking; ‘they reflect the market’s expectations of the future stream of services associated with the asset’. Moreover, they influence household wealth and thus induce consumer spending and aggregate demand2. Striking is the point made by Lettau and Ludvigson (2001): they imply that 85% of the (post-war) variation in the growth of household net worth is ephemeral; attributable to variations in the stock market wealth component.
‘Periodic sharp sustained increases and then reversals in asset prices lead many to posit irrational price bubbles’ (Hendershott et al, 2004: abstract). Kindleberger (2000:16) describes a bubble as ‘an upward price movement over an extended range that then implodes’.
Although it seems indisputable, among economists there has been debate on whether bubbles actually exist. Cecchetti (2006) states that opponents of the bubble view base their arguments on the logic of efficient markets. This logic is based on the idea that markets include all information available and thus this will automatically eliminates bubbles. Cecchetti however points at the many circumstances this argument fails. The logic of efficient markets is based on arbitrage, and when this fails, market efficiency fails. In reality, even when no-one denies there is a bubble, in many cases arbitrageurs will not eliminate the bubble. LeRoy (2004) underlines this belief by stating that there is a strong presumption that bubbles actually exist.
Given that bubbles occur, should a central bank deflate a bubble? Should monetary policy be used to prevent a bubble from busting? Janet L. Yellen (2005)3 stated the following: ‘In answer to the third question on whether monetary policy is the best tool to deflate a house-price bubble, there are several points to consider. For one thing, no one can predict exactly how much tightening would be needed, or by exactly how much the bubble should be reduced. Beyond that, a tighter policy to deflate a housing bubble could impose substantial costs on other sectors of the economy that would lead to equally unwelcome imbalances. Finally, it's possible that other strategies, such as tighter supervision or changes in financial regulation, would not only be more tailored to the problem, but also less costly to the economy.’ Posen (2006) states that central banks should not even try to burst bubbles. Properly supervised and regulated financial system will not suffer much in real terms from irrational increasing asset prices possibly followed by a bust. On the other hand, in countries
2
In a speech delivered at the European Economics and Financial Centre, July 11, 2007 3
where the financial system is improperly supervised or fragile, a monetary tightening will likely cost the country more in real economic terms and can not be a replacement for dealing with the underlying financial problems directly. The thoughts of Posen on this subject give rise to some additional questions: how could an economy not be affected by a bursting of a bubble, given the knowledge and (negative) experience we now have concerning bursting asset price bubbles? Why is it not possible to deflate a bubble and remove the potential damage a bubble can cause in the future? Roubini (2006) supports the view of dealing with bubbles pro-actively and states that moderately adjusting the interest rate can influence the development of a bubble and can therefore reduce the economic distortions caused by a bubble. Responding to asset prices however can cause the economy to slow down. This is why often people do not complain or halt the economy during the development of bubbles; the economy grows heavily during these episodes, which seems to be good for a country. Therefore, adjusting the interest rate to a higher level to prevent the country from turning into a recession or economic decline in the future, might not always be the most popular decision for a Central Bank. Maybe, a Central Bank does not have to make this decision, based on the striking example of Ranciere et al (2003:3): ‘Thailand and India are contrasting examples of a steep but crisis prone growth path and a slow but safe growth path. Thailand has experienced lending booms and crises, while India has pursued a safe growth path for credit. GDP per capita grew by only 99% between 1980 and 2001 in India, whereas Thailand’s GDP per capita grew by 148%, despite having experienced a major crisis.’ In other words, booms and crises might not even be the worst scenario for a country in the long run. It can cause a faster GDP growth track and thus increase welfare. This example does leave out one important issue; an increase in GDP per capita does not necessarily mean everyone profits from it. When the wealth is redistributed and the rich become richer, but the poor do not profit from the increase in welfare, still a lot of people will suffer from the burst of a bubble. Unless a government or central bank gives special attention to the poor, deflating a bubble might not be the worst idea. But, is this even possible?
1.1 Main research question
The main research question addressed in this paper is:
‘What is the influence of the short-term interest rate on the development of property prices?’
2. Literature review
This chapter describes the main literature on the subject of asset prices and monetary policy. Paragraph 1 describes the role of asset prices in monetary policy, paragraph 2 presents the literature in favour and against responding to an asset price bubble, paragraph 3 discusses the various monetary policy reaction functions developed over time, paragraph 4 provides the reader with information on previous empirical studies regarding periods of financial instability and last, paragraph 5 will discuss the literature on asset prices and its response on the short-term interest rate.
2.1 The role of asset prices in monetary policy
‘In the wake of financial liberalization and innovation, asset prices have become more important factors in driving economic fluctuations, allocating resources across sectors and time and influencing the strength of the financial system (Hordahl and Pakker, 2007:1).’
In order to analyze a monetary policy in the light of asset prices, we first need to understand asset pricing. As Bernanke and Gertler (1999) point out, asset price fluctuations basically reflect changes in underlying economic fundamentals. This is the case though, in a world with no regulatory distortions and with efficient capital markets. Borio and Lowe (2002) state that in the early 1990s asset prices seemed to be disconnected from economists. They describe that on the one hand, asset prices stood out in historical accounts of financial instability. These accounts mainly discussed property prices, and sometimes equity prices. On the other hand, studies examining the link between the macro economy and asset prices, focused primarily and almost exclusively on equity prices. Real estate and commercial property accounts were rarely taken into account. Nowadays, the importance of asset prices in monetary policy has gained a lot more attention. Numerous scholars have studied the relationship and the extent to which central banks should incorporate these asset prices in its policy.
Eichengreen and Arteta (2000) found that a 1-percentage point increase in the rate of domestic credit growth increases the likelihood of a banking crisis in the next year by 0.056 percent. This study and others conclude that an increase in domestic credit implies an increase in asset prices, which might be a predictor for an asset price boom followed by a banking crisis.
Asset prices play various related roles in the monetary policy/financial stability frameworks. The last decade, policymakers had to deal with abnormal fluctuations in asset prices. This included strong boom and bust periods, unusually strong rising residential property prices, exceptionally low long-term interest rates, low volatility and narrow credit spreads
(Hordahl and Pakker, 2007).
Bernanke & Gertler (1999) elaborate on two conditions influencing asset price volatility and which should alert central banks. Sometimes nonfundamental factors underlie asset market volatility; changes in asset prices unrelated to fundamental factors (such as the return volatility of the underlying asset, futures return on volatility on the asset, call and put option implied volatilities over various maturities, exercise prices etc.4) have potentially significant impacts on the rest of the economy. Sources of nonfundamental fluctuations in asset prices are for example poor regulatory practice and imperfect rationality on the part of investors5. The second condition necessary for asset price volatility of concern to policymakers is that boom-bust cycles in asset markets have important effects on the real economy. Although it is complicated to learn the real effects of asset price fluctuations, according to Bernanke and Gertler, they can have important effects on the economy. Mishkin (2008) refutes this last statement though by stating that ‘the macroeconomic consequences of asset price fluctuations are unlikely to have long-lasting consequences for the economy as long as monetary policy responds appropriately6‘.
4
Hwang, S. and Satchel, S.E. (2000) in ‘Market risk and the concept of fundamental
volatility: Measuring volatility across asset and derivative markets and testing for the impact of derivatives markets on financial markets.’
5
AllBusiness (1999), Monetary Policy and Asset Price Volatility, By Mark Gertler,
http://www.allbusiness.com/north-america/united-states-missouri-metro-areas/371516-1.html 6
Federal Reserve, (2008), Speech by Governor Frederic S. Mishkin: How Should We Respond to Asset Price Bubbles?
In the past two decades price stability has been the official primary objective of central banks across the world. Stabilization of output volatility or the reduction of unemployment has become a secondary objective. The central bank’s policy tool nowadays is the interest rate, even though for many decades quantity theorists like Milton Friedman and more recently Robert Lucas encouraged the control on money growth as primary rule for monetary policy (Adema and Sterken, 2005).
A general belief nowadays is that extensive central bank independence and a clear mandate for a central bank to control inflation are essential institutional devices to ensure price stability. This is the case at the ECB, but in countries with a more dependent central bank, other considerations, like re-election perspectives of politicians or low levels of unemployment, may hinder the objective of price stability (Sauer, 2007).
How a central bank controls inflation is quite variable. Since the world is globalizing and so is international competitiveness, the model used by a central bank is subject to a lot of uncertainty and variables. As Greenspan described: ‘Despite extensive efforts to capture and quantify what we perceive as the key macroeconomic relationships, our knowledge about many critical linkages is far from complete and, in all likelihood, will remain so. Every model, no matter how detailed or how well conceived, designed, and implemented, is a vastly simplified representation of the world ... Given our inevitably incomplete knowledge about key structural aspects of an ever-changing economy and the sometimes asymmetric costs or benefits of particular outcomes, the paradigm on which we have settled has come to involve … risk management7.’
The extent to which a central bank targets inflation differs; one is more precise in targeting a certain level of inflation then the other. Though, as well as Greenspan, King (2004)8 points out that nowadays behaviour of central banks is quite identical. The main difference is that not all inflation targeting countries are willing to be unambiguous about the long-run inflation rate that comprises price stability.
7
Economist’s View (2005), Greenspan sum’s up,
http://economistsview.typepad.com/economistsview/2005/08/greenspan_sums_.html 8
Asset price bubbles create multiple distortions in a country’s economy; its consumption, investment, the path of fiscal policy, pensions fund management, insurance companies and the balance sheets of commercial banks (Cecchetti, 2006). Cecchetti states that bubbles compromise the stabilization objectives (price stability) of central banks. It is interesting however, that central banks still claim not to be willing, based on several arguments discussed in this paper, to target asset price inflation. Even though as stated by Mishkin (2008)9 ‘the influences of asset prices on demand and inflation through traditional wealth and cost-of-capital channels fall directly within the traditional concerns of monetary policy’. In that same speech, despite of this strong argument, he states that monetary policy should only respond to changes in the outlook for inflation and aggregate demand possibly resulting from asset price movements, rather then ‘pricking’ a bubble. But if asset price inflation leads to CPI inflation eventually, why not responding to it right away? Why waiting till maybe some damage has occurred already? Even though enough arguments lay the foundations to target asset prices, Central Banks seem to be unwilling to respond when asset prices inflate too radically.
Galbraith et al. (2007) claim to be able to explain this response of several Federal Bank Officials. Since except for the growth and cost issue, a Central Bank faces a distribution issue. In this article Galbraith et al. discuss the role of the Federal Reserve concerning unemployment and inflation. Because how does the Fed balance between favoring those who hold assets and those who gain their income through labor?
In line with this ‘balancing-act’ John B. Taylor developed a simple monetary policy rule in 1993. This rule is applied on the behaviour of several Central Banks, including the Federal Reserve and the European Central Bank (Sauer, 2007). The rule sets the quarterly average level of the short-term policy interest rate (it) in response to (four-quarter) inflation (πt) and the output gap (yt), where r* is the equilibrium real rate and π* is the inflation target10:
it= r* + πt + 0,5(πt- π*) + 0,5 yt.
9
Federal Reserve, (2008), Speech by Governor Frederic S. Mishkin: How Should We Respond to Asset Price Bubbles?
www.federalreserve.gov/newsevents/speech/mishkin20080515a.htm#f8 10
This rule states that a Central Bank should increase the interest rate when inflation is above and unemployment is below their targets, vice versa it should decrease the interest rate when inflation is lower then and unemployment is higher then their targets. If only one of the two variables is below or above its target, a Central Bank should not act since in that case the signals from the two variables conflict according to Galbraith et al. (2007). Other authors have explained this rule differently. For example Adema and Sterken (2005) stated that when inflation increases according to the expectations, the interest rate has to increase more than the increase in inflationary pressure. This act will increase the expected real interest rate and may result in lower expected inflation. To summarize, they only demanded one of the two variables to fall below or above its target.
To research whether or not the Federal Reserve favours either the labour force or those who hold assets, Galbraith et al. researched when the Fed adjusted its interest rates. They found striking evidence that the Federal Reserve reacts to low unemployment by increasing the interest rate, but when unemployment is above its target it does not necessarily decreases its interest rate. On the other hand, since 1983 the Fed does not respond to high or rising inflation by tightening the interest rate nor does it ease monetary policy when inflation is below target. When inflation was above target and unemployment below the central bank did act though.
This article seems to shed some light on the issue; when responding to inflation a central bank favours those who hold assets and neglects the labour force. Consciously or unconsciously a central bank has to choose between those two groups. Since the assumptions of the Taylor Rule that both unemployment and inflation rate shift at the same rate in the opposite direction, does not happen all the time.
2.2 Arguments in favor and against responding to an asset price bubble.
Although for many years, the dominant idea was that a central bank should stay away from influencing the development of asset prices, lately, many scholars have challenged this view.
some have stated that in the United States, the Federal Reserve kept the bubble going, by injecting vast amounts of liquidity into the markets to prevent the bubble from bursting too heavily. Catherine Austin Fits (2004)11stated during an interview in 2004, that at that moment Federal agencies, mortgage agencies and the U.S. Treasury were financing a political economy: ‘You can keep a bubble going as long as you can finance it’, according to Austin Fits. Question remains which policy limits the costs the most?
Since every asset price boom is different and every country faces other economic conditions, the answer to this question is not necessarily the same in every case. Bordo and Jeanne (2002:3) state that ‘to be optimal, a proactive monetary policy must come into play at a time when the risk is perceived as sufficiently large but the authorities’ ability to act is not too diminished’. From this statement, uncertainty can be read. When is a bubble large enough?
Both Bernanke and Gertler (1999) and Hessius (1999)12state that fluctuating asset prices are caused by both fundamental and nonfundamental factors; if a bubble is identified it is still very hard to burst it without damaging financial markets and the rest of the economy. Posen (2006) does not agree with this argument. He agrees with Blanchard (2000), stating that central banks continuously make decisions based on uncertain and difficult assessments of indirect observable variables such as potential output and equilibrium exchange rates. But Posen (2006:110) does agree not to prick bubbles, he states; ‘(…) it is simply not worth it. Attempting to do so using monetary instruments will almost certainly fail, because the connection between monetary conditions and asset markets is far less tight than commentators assert. ‘
Cecchetti et al. (2000) on the other hand, motivate a pro-active central bank policy by pointing out that this will likely reduce the chance of asset price misalignments, that is, in the normal course of policy-making. Moreover, forecasts about the inflation depend on assumptions about asset prices. These assumptions in return, must depend on visions about the extent of asset price misalignments. They do not however recommend that central banks should respond to every asset price bubble, to target specific levels of asset price, or to react to every asset price misalignment or bubble in the same way. Roubini (2006:89) takes it one
11
Pete McCormack (2004), The Real Economy in Fitts and Starts, Catherine Austin Fitts on Economic Warfare and Trillions of Missing Dollars,
www.petemmccormack.com/CAFinterview-footnotes.pdf 12
step further and does encourage asset price targeting, he states: ‘Thus, optimal monetary policy should pre-emptively deal with asset bubbles rather than just mop up the mess that they cause after they burst.’ In order to design an optimal monetary policy though, a reaction function should be able to distinguish between the fundamental and nonfundamental component of the asset price. More important, as Jean-Claude Trichet13 points out: ‘(…) the challenge to call a boom a bubble is of another order of magnitude if the judgment has to be made in real time, which means when the boom is developing.’
Responding or not responding, or keeping a bubble ‘alive’; with every policy costs are involved. The optimal monetary policy should indeed limit the costs as much as possible. Unfortunately though, and backed by the amount of literature written on this subject, one optimal monetary policy has not been found yet, and probably never will be found. Since, every country, economy, period in time and situation is different and therefore difficult to generalize. Still, in the following chapter an overview can be read on many scholars who have attempted to develop this optimal monetary policy reaction function.
2.3 Monetary policy reaction functions
When analyzing the arguments in favor and against using monetary policy to target asset prices, this brings us to the monetary policy rules or the reaction function of a monetary authority. Should asset prices be incorporated in such a reaction function? Roubini (2006) states that they should when the goal is to achieve an optimal monetary policy, in addition to the direct effects that asset prices already have on expected inflation and current growth.
Whether asset prices should be incorporated in monetary policy reaction functions has been debated heavily. In 1999, Bernanke and Gertler extended a small-scale macroeconomic model developed by Bernanke, Gertler and Gilchrist (2000), the BGG model. This is based on a standard dynamic New Keynesian model, adjusted to allow for financial accelerator effects. The BGG model assumes that only fundamentals, such as shocks to productivity or spending, drive asset prices. Bernanke and Gertler extend this model, to allow for exogenous bubbles in asset prices14(nonfundamental factors affecting asset prices). They show, that with the use of
13
Speech by Jean-Claude Trichet, President of the ECB, during a Mas lecture on 8 June 2005, Singapore
14
Roubini (2006): ‘The term ‘exogenous bubbles’ refers to bubbles where the bubble
this model, in the case of an aggressive inflation targeting policy, an asset bubble has moderate effects on the output gap and inflation. In the case of an asset price targeting monetary policy, the variance of output and inflation is larger. In the moderate approach, output and inflation decline, in the aggressive approach, inflation is stable though the output gap shows a larger variance. The main criticism on the results of Bernanke and Gertler is that the policy reaction function to the monetary authorities is not ‘optimal’. The size of the coefficient of interest rates on output and stock prices and expected inflation is chosen in a rather ad hoc way (Roubini, 2006).
Cecchetti et al.(2000) state, contrary to the claims made by Bernanke and Gertler, that the central bank should react systematically to misalignments in asset prices, on top of their reaction to inflation forecasts and output gaps. They based their arguments on three simulation experiments; (1) a closed-economy model subject to a stock price bubble, (2) monetary policy and the exchange rate in a small model of an open economy and (3) reacting to the exchange rate in a large multi-country model. According to Cecchetti et al., asset price bubbles cause distortions in investment and consumption, which can lead to dramatic increases followed by falls in real output and inflation. By tightening monetary policy modestly when asset prices rise above target levels and easening monetary policy when asset prices fall below target levels, this can moderate the impact on output and inflation and enhance overall macroeconomic stability.
In the former models, Cecchetti et al. and Bernanke and Gertler consider exogenous deterministic bubbles. Bernanke and Gertler (2001) respond to Cecchetti et al. by modeling the non-fundamental component of their original model as an exogenous stochastic process. Moreover, they increased the elasticity of Tobin’s q15 with respect to investment with 1.5, which is consistent with the evidence. Again the results show that with an aggressive inflation targeting policy inflation variability and output gaps are reduced. When monetary policy responds to stock prices the output gap variability reduces, but as Bernanke and Gertler state, this will likely be outweighed by the associated increase in inflation variability.
Deterministic bubbles refer to cases where the bubble is modelled with no uncertainty, i.e. the time period during which the bubble continues and the time at which the bubble bursts are known to all with a probability of one.’
15
With these results Bernanke and Gertler showed that their model holds in case of stochastic bubbles, but they were not able to refute the criticism of Cecchetti et al. on their ad hoc coefficients, since they did not optimize them in this research either.
Filardo (2000) did a research on incorporating asset prices (not specifically asset price bubbles) in monetary policy, using a fairly simple macroeconomic model (based on a model of Ball (1999) and Rudebusch and Svenson (1999)). The model has two main components: an equation describing monetary policy and a model of CPI inflation and output. Housing price inflation enters the model twice; (1) it is included in the CPI inflation equation because property prices are assumed to help predict CPI inflation in the future and (2) it enters the monetary policy component of the model because central banks might find it useful to respond to changes in asset price inflation. His paper recommends that in a situation without uncertainty about the role of asset prices in determining inflation and output the optimal monetary policy would be to react to asset prices. In the case of uncertainty about macroeconomic consequences of asset prices, a monetary authority would not benefit from a reaction. The corresponding reason is that housing price inflation shows some power to predict future CPI inflation, but stock market price inflation does not. Therefore, using asset prices to improve optimal monetary policy is not promising, according to Filardo. In 2001, Filardo extends the model of Filardo (2000) and adds a role for asset price inflation (for both the fundamental and bubble component). The model contains three modeling blocks: the monetary policy block, the asset price inflation block and the macroeconomic block. Filardo now concludes that monetary authorities should respond to changes in both general asset prices and asset prices bubbles. When the bubble increases monetary policy should be tighter than under the Taylor rule, but when it bursts, monetary policy should be easier than under the Taylor rule. Even in the case of uncertainty about identification of the bubble component, asset prices can still hold valuable information.
Two important remarks are made by Dupor (2002) and Smets (1997), and these are important to keep in mind when discussing the role of asset-prices in monetary policy.
labor supply and accordingly a fall in inflation. Consequently, with only one instrument, being the short-term interest rate, stabilizing both asset-prices and inflation cannot be achieved at the same time, this is also known as the Tinbergen rule.
Second, Smets notes that the source of asset price movements influences the way policy makers respond to these movements when attempting to maintain price stability. For instance, when equity prices rise because of nonfundamental shocks in the equity market, optimal policy will require a raise in interest rates. But, when equity prices rice because of fundamental shocks, like shifts in total factor productivity, central banks may want to accommodate the boom by keeping the real interest rate untouched.
This brings us to the question: ‘Is it even possible to develop an optimal monetary policy reaction function which incorporates asset prices, but at the same time distinguishes exactly between fundamental and nonfundamental asset price rises?’ Or should central banks analyze every asset price boom separately and respond (re-active or pro-active) accordingly? Summarizing the above arguments, this research will not create an optimal monetary policy reaction function in which the asset price is incorporated. At this point in time, knowledge about the components of asset prices and asset price booms is simply too limited to develop a sensible and useable rule.
2.4 Empirical research
In the previous paragraph a brief overview of theoretical approaches has been described. This paragraph provides the reader with information on previous empirical studies regarding periods of financial instability.
Smets did, but did have quarterly instead of annual data. By means of VAR Technology, Detken and Adalid manage to confirm their hypothesis that there are significant differences between high- and low-cost boom periods. A Vector Autoregression (VAR) model is the extension of the autoregressive (AR) model when more then one variable is studied. In an AR model the independent variables are lags of the dependent variable (Koop, 2005). The depth of post-boom recessions are explained by factors like for example money growth shocks and real residential property price developments even when one controls for housing price developments and the monetary policy stance, which is measured by means of divergences from a Taylor-rule during the boom and in the post-boom phase.
2.5 Short-term interest rates and asset prices
The above stated literature review discusses whether or not a central bank should interfere in asset price booms based on empirical and theoretical research. One important question remains unanswered though. This thesis will research whether a monetary policy can control asset price growth or decline. Is the Central Bank’s main tool, the short-term interest rate, influential enough to stop deviations from asset price trend? The central bank controls short-term interest rates, but does this mean that when these interest rates are cut, bank and market interest rates decrease as well and asset price growth increases? Does an increase in the interest rate mean that asset price growth declines? Is the transmission mechanism as presented by the European Central Bank really that effective when controlling asset prices?
2.5.1 Equity Prices
Rigobon and Sack (2004) discuss the relationship between equity prices and the short-term interest rate and point at two difficulties when analyzing this relationship. First short-term interest rates are concurrently influenced by movements in equity prices. This results in a difficult endogeneity problem. Second, other variables, like information about the macroeconomic outlook or changes in risk preferences, also have an impact on both short-term interest rates and asset prices. Rigobon and Sack estimate the response of asset prices to monetary policy by employing a technique called ‘identification through heteroskedasticity’. The approach focuses on changes in the co-movements of asset prices and interest rates, when ‘the variance of one of the shocks in the system is known to shift’. Using this technique the response of asset price to changes in monetary policy can be estimated under rather weak assumptions. The results show that an increase in the short-term interest rate has a negative impact on stock prices. Gürkaynak, Sack and Swanson (2004) state that the effect of monetary policy on asset prices is characterized by two factors. These two factors are the ‘current federal funds rate target’ and a ‘future path of policy’ factor. The current federal funds rate target corresponds to surprise changes in the current federal funds rate. The future path of policy factor corresponds to changes in the future rates of up until one year that are independent of changes in the current funds rate target. Their analysis shows that Federal Open Market Committee (FOMC) statements of the Federal Reserve have significant impact on asset prices. It can be suggested that ‘the FOMC has the ability to conduct policy with a substantial degree of commitment to a state-contingent, or conditional, path for the funds rate several quarters or even years to the future (p. 21)’. This implies that the FOMC is largely unimpeded in its ability to conduct policy: it has the ability to influence financial market expectations of policy actions in the future. This implies that the FOMC can influence longer-term interest rates and thus the economy in general.
The above-discussed papers, analyze the influence of the short-term interest rate, or federal funds rate, on asset price growth on an intraday basis. In this thesis, the longer-term effects of a change in the short-term interest rate to asset prices will be analyzed. Therefore, the information about the macroeconomic outlook or as mentioned in the article of Gurkaynak et al. (2004), FOMC statements, will not be taken into account.
2.5.2 Property prices
In a study by Mésonnier and Lecat (2005) based on a panel of 18 industrialized countries, different potential determinants16 of property prices are estimated. They use the panel regression technique, which utilizes both the periodic and cross-section dimension of a database containing several observations. The equation is estimated using the Generalised Method of Moments (GMM), as applied on dynamic panel data by Arellano and Bond (1991). The advantage of this technique is that it able to deal with possible endogeneity problems of the explanatory variables. This econometric analysis confirms the impact of both short- and long-term interest rates on property price growth.
The IMF (2004) World Economic Outlook uses the same technique (GMM) as Mésonnier and Lecat to analyse fluctuations in residential property prices in 18 countries. The research analyzes the extent to which fundamentals explain fast increase in house prices. It builds on and extends a dynamic panel model developed by Lamont and Stein (1999). This model assumes that the growth rate of real house prices is explained by (1) the past growth rates of real house prices, (2) the past housing affordability ratio and (3) economic fundamentals. The econometric results in the IMF paper show that real house prices show high persistence, long-run reversion to fundamentals and dependence on economic fundamentals. Moreover, there seems to be a strong tendency for real house prices to rise tomorrow if they rise today (growth rate has a serial correlation coefficient of 0.5). More importantly for my research, the results show that a 1 percent increase in the short-term interest rate results in a 0,5 percent decrease in real house price inflation. Borio and McGuire (2004) analyze peaks in equity and housing prices in 13 industrialised countries from 1970 Q1 to 1999 Q4. They discover that there is a clear tendency for peak in equity prices to precede peaks in housing prices. Analyzing this tendency, they research whether equity prices help to predict peaks in housing prices (a) on their own, (b) after allowing for the macroeconomic variables that traditionally have been found to clarify housing price movements (the control variables) and (c) after allowing for financial imbalances built up during the boom. The predictive power of the different variables is determined through a series of probit regressions, in which all control regressors include single lags of GDP growth, changes in short-term nominal interest rates and changes in unemployment. They find that equity price peaks tend to precede housing price peaks with a lag of at least one year. This was particularly the case when the house price peaks were preceded by the development of financial
16
3. Empirical Research
To research the influence of the short-term interest rate property price development, I use real property price indices for 18 OECD17 countries and corresponding selected macro-economic and financial variables1819.
3.1 Economic model
In this research I will seek to assess the extent to which the short-term interest rate explains the development of property prices, by specifying the following general equation:
5 2
GRRPPi,t=
∑
βaGRRPPi,t-a+ β6∆ SRi,t+ β7∆LRi,t +∑
γaGRGDPi,t-a +a=1 a=1
2
+ γ3GRPCRi,t-1 +
∑
γ4GRGREPi,t-c+ γ5BCi,t+ γ6BQi,t+ c=1γ7(VARi,t* ∆SRi,t)+ µi + θt + εi,t
i is the country index = 1,…..,18 and
t is the time dimension = 1971Q3,……, 2004Q3 (not starting at 1971Q3, where my dataset starts, because of lagged variables) and t= 1985Q1, …..2004Q3
The dependent variable GRRP is Growth Rate Property Prices. As a first regressor I have included 5 lags of the dependent variable of Property Price Growth Rate to capture the dynamics of housing prices and to control for initial conditions20. The second regressor (∆SR) is the change in short-term interest rate, and the third regressor (∆LR) is the change in long-term interest rate which both are expected to have a negative sign. Aside from being particularly interested in the significance and sign of the my main variable of interest, the short-term interest rate, additional special interest goes out to the long-term interest rate. The motivation behind this is that, as stated by Mésonnier and Lecat (2005), Sutton (2002) and Tsatsaronis and Zhu, there are both countries where housing loans are mainly based on
short-17
Sample specification can be found in the appendix, table A2. 18
Many thanks to Carsten Detken and Ramón Adalid for providing me their dataset, the variable definition and their source can be found in the appendix, table A1.
19
The variables used are real variables, i.e. are adjusted for the differences in price level (inflation).
20
term markets rates and countries where ‘long-term interest reflect the rate used to discount the housing service flow’. µiand θt are country and time dummies, and εi,t is the error term. The control variables I use are chosen based on previous literature and their expected significance in determining asset price growth.
A variable for the GDP growth rate (GRGDP) will be included, which is a measurement of the state of the business cycle and household income, a variable that has traditionally been found to explain housing price movements (Borio and McGuire, 2004). This variable also summarizes the information contained in more direct measures of household income, such as unemployment, wages and population growth. Following Tsatsaronis and Zhu (2004) these latter variables will not be included on grounds of parsimony of specification.
In line with Mésonnier and Lecat (2005), IMF (2004) and Tsatsaronis and Zhu (2004) a variable for private credit growth rate (GRPCR) will be included, (which is a proxy for mortgage debt, and given the imperfections of the credit market, a useful supplement to describe asset price growth not explained by interest rates). Moreover, a variable to control real for stock price growth (GRGREP) will be included, because research has shown that this does have a significant influence on house price growth (McCarthy and Peach, 2002).
Last, two intercept (value 0 or 1) and a slope (or interaction) dummy variable will enter the equation; a dummy variable representing a banking crisis (BC), a dummy variable representing a boom episode (BQ), and a dummy variable representing the interaction between the short-term interest rate and countries with a variable mortgage rate system. A banking crisis is typically associated with a drop in asset prices (following Mésonnier and Lecat, 2005)21. The reason for including a episode dummy variable is that boom-episodes cause outliers in property prices and thus have a significant effect on the regression22. The slope dummy variable for countries with a flexible mortgage system is included, because it is expected that countries with flexible mortgage systems are more sensitive to property price changes when the interest rate changes then countries with fixed mortgage systems (following Tsatsaronis and Zhu, 2004)23.
21
Specified table with banking crisis episodes can be found in the appendix, table A5. 22
The boom-episodes are based on the boom-episodes estimated by Adalid and Detken (2007) who used the same dataset as I do. An overview of asset price boom episodes can be found in the appendix.
23
Hypothesis:
An increase (decrease) in the short-term interest rate has, all other variables taken into account, a negative (positive) effect on real property price growth.
The hypothesis will be tested for two time episodes; 1971Q3 to 2004Q3 and 1985Q1 to 2004Q3. In the article by Mésonnier and Lecat (2005) it was stated that during the 80s large deregulations of financial markets occurred and these deregulations put pressure on lending rates, fostered financial innovation and caused the financial environment to transform. These deregulations were often followed by a sharp increase in real property price inflation and credit booms. Kingdom, Muellbauer and Murphy (1997)24 found that after this deregulation interest rates and income expectations became more important determinants of property price growth. This is the reason I will test for both the whole sample period, and the sample starting in 1985Q1, the time period where most deregulations had started25. My expectations in terms of coefficient signs, do not change, however, I expect that the value of the interest rate coefficients will be higher.
Figure 2 - Overview of Explanatory Variables - Expected signs
Variable Expected Sign
Log difference of Real Property Price β > 0
Change Short Term Interest Rate β < 0
Change Long Term Interest Rate β < 0
Log difference of Real Equity Price γ> 0
Log difference of Real Private Credit γ> 0
Log difference of Real GDP Growth γ> 0
Dummy Boom Quarter γ> 0
Dummy Bank Crisis γ < 0
Slope Dummy Variable∆SR*VAR γ < 0
3.2 Methodology used for research
The model used for this research needs to be able to cope with some facts. Since other variables, also cause the development of property price growth, the model used needs to control for these effects while focussing on my main variable of interest; the short-term interest rate. The fixed effects model could be an appropriate instrument. But, as a lagged
24
Referenced from article Mésonnier and Lecat (2005) 25
dependent variable is included in the regression analysis, this would possibly cause biased results; since the disturbance of the lagged dependent variable is included in time mean εi,t it suggests a non-zero correlation between the error term and the lagged dependent variable. This was stated by Nickell in 1981, where the presented results showed that the Least Square Dummy Variable (Fixed Effects) estimator is not consistent in dynamic panel data models. One important addition has to be made though; Nickell states that the bias produced by LSDV is O(1/T); as T grows, this bias becomes smaller (Gaduh, 2002; Judson and Owen, 1999). Various scholars came up with solutions based on Instrumental Variable and Generalized Method of Moments Estimators. The most discussed and used estimators, are the ones produces by Anderson and Hsiao (1982), Arellano and Bond (1991), and Blundell and Bond (1998).
3.3 The Least Squares Dummy Variable Method or Fixed Effects model
The technique used to regress the equation, will have to cope with a number of facts; (i) Time period (T) starts 1971 Q3 and ends 2004Q3; T=129
(ii) Time period (T) starts 1985Q1 and ends 2004Q3; T=79 (iii) The number of countries (N) is 18
(iv) The panel data set is unbalanced
(v) The right-hand side of the equation has lagged dependent variables and lagged independent variables.
In retrospect: the previously presented most used estimators (Anderson-Hsiao, Arellano-Bond, Blundell-Bond) all have one main short-coming: the properties of these estimators hold with a large number of cross-sectional units, but when this is not the case the results can be severely biased and inaccurate (Bruno, 2005). In my research with 18 countries, this can cause serious problems.
4. Testing the assumptions
Before presenting the results from the regressions, several tests have to be performed. In this paragraph the tests will be presented and the results will be discussed.
4.1. Descriptive statistics and normal distribution
When looking at the descriptive statistics for both sample sizes, it is noticeable from the standard deviations of the variables and the Jarque Bera test that my series are not normally distributed. To improve the situation outliers are studied using box-plots to identify and remove the far outliers. This will prevent the results from being driven by these outliers. However, not all outliers will be removed since macro-economic data in general are not normally distributed. As stated by Startz (2007) and Hill et al. (2001), there is no requirement that variables in a regression are normally distributed when the sample size is sufficiently large. In that case the least squares estimators have a distribution that approximates normal distribution. Since the dataset is large, near outliers will not be removed to attempt to reach the goal of normal distribution; this will not make the results more realistic26. The descriptive statistics without far outliers present similar results compared to the descriptive statistics with outliers; the standard deviations however have decreased. When analyzing the descriptive statistics, it is interesting to note that the means of the change in interest rate are always negative. This means that the interest rate is more often adjusted downward then upwards. Does this imply that inflation has been less of an issue then unemployment?
4.2 Collinearity and Multicollinearity
Collinearity exists when variables correlate with each other (they move together in systematic ways). Multicollinearity exists when collinearity arises between a number of explanatory variables. To test for collinearity, a correlation matrix is used; this correlation matrix shows no reasons to worry about collinearity in both sample sizes. The Variance Inflation Factor method tests for multicollinearity show that no problematic multicollinearity exists27.
4.3 Stationarity
Stationarity implies a stochastic process, where the probability distribution is the same for all times. Therefore, mean and variances do not change over time. When the mean and variances do change over time nonstationarity exists. To test for stationarity in panel data, the generally
26
Descriptive statistics can be found in the appendix table B1, B2, B3 and B4 27
used Dickey-Fuller, Augmented Dickey-Fuller and Phillips-Perron tests cannot be used, because these tests do not have the power to differentiate the unit root null from stationary alternatives (Maddala and Wu, 1999). As an alternative, the Levin-Lin-Chu (LLC) and Im-Pesaran-Shin (IPS) tests are widely used. Both tests have their shortcomings though.
Maddala and Wu (1999) state that the LLC test does wrongly fail to reject the null-hypothesis of a unit root occasionally and is therefore not always reliable. The IPS test on the other hand only considers balanced panel data; the dataset used in this research is unbalanced. Moreover, both tests assume that T and N go to infinity; the number of cross-sections in this dataset is limited to 18.
In general, there has been quite some discussion on the topic of unit roots in panel data. (Libanio, 2005) for example states that there does not seem to be a clear consensus about the appropriate methodologies to perform unit root tests, the theoretical importance of the concept of unit roots and its implications for macroeconomic analysis. Moreover, Banjeree (1999) points out that when cross-section and time series are pooled, this might decrease the unit root effects.
Since both the IPS and LLC tests have their shortcomings and do not ‘fit’ the dataset used in this research, running these tests on my dataset might not give reliable results. Moreover, the theoretical importance and real implications of unit roots in macroeconomic data is heavily debated. Therefore, the results of the tests can be found in the appendix, but will not further be discussed in this research.
4.4 Heteroskedasticity
4.5 Autocorrelation
Autocorrelation violates the OLS assumption that the error terms are uncorrelated. To test for autocorrelation, the Durbin H test is employed. Since lagged values of the dependent variable are used amongst the regressors, the common Durbin Watson statistic is an unreliable measure of autocorrelation (Watson, Teelucksingh, 2002). Therefore, instead of reporting the Durbin Watson statistic, I will report the Durbin H statistic28.
4.6 Reverse Causality
Last, reverse causality of the variables in the regression is tested for. The development of property prices can be influenced by the interest rate, but the interest rate can also be influenced by the development of property prices (as is argued by those in favor of responding to asset prices). To test for reverse causality the Granger Causality Test was employed. Indeed, this showed evidence of reverse causality of the interest rates and housing investment variable. Reverse causality can be a problem and can cause biased results.
A possible solution to this problem could be to use the GMM/Instrumental Variables approach; though as stated by Judson and Owen (1999) and Gaduh (2002) these approaches produces biased estimates when T is large. Another solution would be employing a completely different model, namely the VAR model, which was used by Tsatsaronis and Zhu (2004); this model treats all variables as endogenous. Replacing the change in short- and long-term interest rate by there primary value, the interest rate, does not solve the potential problem of reverse causality. A last solution to this problem could be to lag the interest rates; however as estimated by the Schwartz information criterion this would not be optimal. Therefore, the possible biased results caused by reverse causality are approached as a possible limitation to my research.
28
5. Results
This chapter will discuss the results of the regressions. In paragraph 5.1 the results of the larger sample (1971Q3-2004Q3) will be discussed. In paragraph 5.2 the results of the smaller sample (1985Q1-2004Q3) will be presented. First the fixed effects likelihood test is performed to check the significance of cross-section and period fixed effects for both sample sizes:
Figure 3 – Results Fixed Effects Likelihood Test Redundant Fixed Effects Test
Results Sample 1971Q3-2004Q3 (probability) Sample 1985Q1-2004Q3 (probability) Cross-section F 0.3827 0.4053 Cross-section Chi-Square 0.2930 0.3086 Period F 0.0000 0.0000 Period Chi-Square 0.0000 0.0000 Cross-section/Period F 0.0000 0.0000 Cross-section/Period Chi-Square 0.0000 0.0000
The results prove the significance of both the period and cross-section fixed effects and period fixed effects only. Using cross-section fixed effects individually appears not to be significant. In Appendix C, the output of the regressions can be found. The regressions for both sample sizes were performed using both cross-section and period fixed effects and using period fixed effects only. The results of the F-statistic29 and the adjusted R2(30) show however, that the model using period-fixed effects only is most significant and has the highest goodness of fit (for both sample sizes). Therefore, the results of these regressions will be discussed31.
5.1 Results Sample 1971Q3-2004Q3
In the appendix, table C1, the results of the regressions performed can be found. The first column reports the regression results with both period and cross-section fixed effects, the second column reports the regression results with period fixed effects only. The latter model will be discussed for the reasons explained earlier.
29
The F-test ‘tests the overall significance of the model’ (Hill et al, 2001:174) 30
The adjusted R2shows how much the variation in property price growth about its mean is explained by the variations in the explanatory variables (the goodness of fit of the model) 31
Looking at the coefficients of the interest rates shows some interesting facts. First, the coefficient of the change in long-term interest rate is negative and significant as was expected. The coefficients of the change in short-term interest rate and the interaction term show that of the short-term interest rate on property price development is highly insignificant.
When looking at the results of the control variables shows some interesting results. The 5thlag of the property price coefficient is negative and significant; implying that the current property price growth rate is negatively influenced by the property price growth rate 5 quarters earlier in time. Adalid and Detken (2007) also found a negative and significant coefficient on the 5th lagged quarter of property price growth. Moreover, the results show that the 1st lagged quarter equity price growth rate is not significantly influencing property price growth; while the 2nd lagged quarter is. This implies that it takes 2 quarters for the property price to respond to equity price fluctuations. The other control variables all have the expected signs and are significant at a 1%.
Concluding, first of all, the regression results are robust and the Durbin H test shows that the hypothesis of no autocorrelation is accepted, therefore autocorrelation is not a problem in any of the regression results. Moreover, the adjusted R2 shows that 47% of the variation in property price growth is explained by the explanatory variables.
These results imply that in the period from 1971Q3 till 2004Q3 overall, an increase of the long-term interest rate by 1 percentage point, would imply a 0,29 percentage point decrease in property price inflation. The result concerning the insignificance of the short-term interest was not expected. The corresponding reason can be that in the first 15 years of this sample period most countries did not experience deregulation practices. These deregulation practices which started around the mid 1980s caused the property price to be more sensitive to changes in the short-term interest rate. This is supported by the findings of Mésonnier and Lecat (2005) and Kingdom, Muellbauer and Murphy (1997). When taking the whole sample period, the effects of the short-term interest rate are eliminated, while the effects of the long-term interest rate are visible. As a result, the hypothesis is rejected. Therefore, in the next paragraph, the focus will be on the period the deregulation practices started.
5.2 Results Sample 1985Q1-2004Q3
Opposite to the results of the regression with the larger dataset, the coefficient on the long-term interest rate is insignificant, while the coefficient on the short-long-term interest rate is negative and significant. The interaction term though is insignificant; implying that property prices of variable mortgage system countries do not respond to changes in the short-term interest rate stronger.
The control variables again show some unexpected results. The second and fifth lag of property price growth is negative, although it is only significant again for the latter lag. Again in this model, the first lag of equity price growth is insignificant. All other explanatory variables show the expected signs and are significant.
Concluding this paragraph regarding the regression results of the smaller sample size, first of all the regression results are robust. The Durbin H test shows that autocorrelation is not a problem in any of the regression results. The adjusted R2shows that the explanatory variables able to explain 49% of the variation in property prices. The results regarding the interest rate variables show that for this data sample, the long-term interest rate does not significantly influence the property price development. In contrast, the coefficient on the short-term interest rate variable is significant and negative; when the short-term interest rate increases by 1 percentage point, property price inflation decreases by 0.22 percentage point on average32. This is in line with the expectations and is supported by the findings of Mésonnier and Lecat (2005) and Kingdom, Muellbauer and Murphy (1997) as discussed earlier. Therefore, the hypothesis can be accepted for the period from 1985Q1 till 2004Q3.
32
6. Conclusion
The recent mortgage crisis in the United States has fuelled discussion on the role of the Central Bank in the development of asset prices. Many critics state that the Federal Reserve should have responded earlier to the dramatic rise of housing prices, others defended the Fed’s strategy, by stating that responding to asset prices would damage more then it would do good. This research assessed the true influence of a central bank on asset prices, more specifically, the influence of the short-term interest rate on property prices.
The main research question addressed in this paper is:
‘What is the influence of the short-term interest rate on the development of property prices?’
In contrast to the research done on the influence of monetary policy on property price development, this research uses quarterly instead of yearly data and moreover takes into account both the long- and short-term interest rate influence on property price growth33. Moreover, contrary to for example Mésonnier and Lecat (2005) and IMF (2004) using the GMM technique and Tsatsaronis and Zhu (2004) using the VAR estimation technique, the size of the database used for this research allows to use the fixed effects (or least square dummy variables) technique to run the regressions.
The regressions conducted in this research were split in two sample periods; one large sample period covering the time from 1971Q3 till 2004Q3, one smaller and more recent sample period covering the time from 1985Q1 till 2004Q3. From the results of the regressions conducted, it becomes clear that in the past 35 years the influence of the interest rate on property price growth has shifted. The larger sample period shows a significant influence on property price growth of the long-term interest rate, while the shorter and more recent time period shows a significant influence of the short-term interest rate on property price growth in the same quarter. This can be explained by the deregulation of interest rates happening in the mid 1980s in most OECD countries. Property prices of OECD countries with variable mortgage rates do not appear to be particularly more sensitive to changes in the short-term interest rate.
Moreover, the significance of the period-fixed effects in the regressions is interesting especially in the light of recent developments in the property and equity markets. The mortgage and now credit crisis in the United States does affect countries and their economies
33
all over the world today; even seemingly stable economies are affected by the problems that started in the US. Asset prices are not only influenced by inter-country developments, but globalization causes asset prices also to be influenced by the state of the global economy at that period in time.
When now analyzing Figure 1, the Transmission Mechanism as presented by the European Central Bank, this research can be concluded by stating that the official interest rate does indeed have an influence on property prices and thus property price development. Central banks can, and some authors state must, use their power to stabilize not only consumer price inflation, but also asset price inflation.
6.1 Limitations and Future Research
The limitations of this research entail the limited power that the economic model used in this research gives to the central bank. Besides its main tool, the short-term interest rate, a central bank also has a short-term informational function, the macro-economic outlooks, which can be of great influence to property price development34. Future research on this subject, including the dates of macro-economical outlooks, would add to the knowledge on this subject.
Moreover, reverse causality exists between the interest rates and property price growth. Both the short- and long-term interest rate can influence property price developments and vice versa. This possibly causes the results to be biased and the results are therefore open to discussion. Using a VAR estimation technique or GMM technique can eliminate these possible biased results.
Last, in this research an attempt was made to analyze the possible increased sensitivity of property prices to changes in the short-term interest rate in variable mortgage rate system countries. The results however are based on a relatively small database; 10 countries have the fixed system, whereas 8 countries have the flexible system. Therefore, future research can elaborate on differences in property price sensitivity to short-term interest rate changes between countries with flexible versus countries with fixed mortgage systems by using a dataset with more countries.
34
APPENDIX
Monetary Policy and Asset Prices
The influence of the short-term interest rate on property price
growth
Anne M. Nijenhuis S1284282
University of Groningen Faculty of Economics and Business
Groningen
Table A1: Variable definition and sources
Economic concept Units Description and Source
Real Residential Prices Index (Quarterly) Real residential prices, from the Bureau for International Settlements (BIS)
Short-term Interest rate Percentage (Quarterly)
Short-term interest rate, from Organization for Economic Co-operation and Development (OECD), Economic Outlook
Long-term Interest rate Percentage (Quarterly)
Bureau for International Settlements (BIS)
Real Gross Domestic Product
Millions (Quarterly) GDP, volume, market prices from Organization for Economic Co-operation and Development (OECD) Economic Outlook, for Germany before 91, from BIS Real Private Credit Millions (Quarterly) Real Private Credit, from International Financial Statistics
(IFS, IMF)
Real Equity Price Index (Quarterly) Bureau for International Settlements (BIS)
Bank Crisis Dummy variable Bank Crisis Episodes, from Caprio and Klingebiel (2003) Boom-Quarters Dummy variable
(Quarterly)
Asset price boom episodes, from Adalid and Detken (2007)
Variable Mortgage System Country
Dummy Variable Overview from the article by Tsatsaronis and Zhu (2004)
A large part of my dataset (except for the dummy variables bank crisis and variable mortgage system countries) is kindly supplied by Ramon Adalid and Carsten Detken. Therefore the information below given on the dataset is not based on my own findings, but on the information given by Adalid and Detken.
With regard to the private credit series, I used the adjusted credit series dataset from Adalid and Detken. They adjusted this dataset since the original one displayed huge structural breaks. ‘Whenever the IFS documentation signaled a structural break and simultaneously the
TRAMO software indicated a level shift (based on the time series characteristics), we let TRAMO estimate the size of the break and used the (backward) corrected data (Adalid and Detken, 2007).’
Table A2: Sample Specification
Australia (AU) Japan (JP)
Belgium (BE) The Netherlands (NL)
Canada (CA) Norway (NO)
Denmark (DK) New Zealand (NZ)
Finland (FI) Spain (ES)
France (FR) Sweden (SE)
Germany (DE) Switzerland (CH)
Ireland (IE) United Kingdom (GB)
Italy (IT) United States (US)
The sample includes 18 OECD countries, with data from 1970Q1 till 2004Q4.
Table A3: Aggregate asset price booms in selected OECD countries (1970-2004)
Australia 1979Q4 - 1981Q4 1988Q3 - 1990Q1* 2003Q2 - 2004Q4** Belgium 1988Q2 - 1990Q3 1998Q1 - 1999Q3 Canada 1988Q1 - 1990Q1* Denmark 1983Q4 - 1986Q4* 1997Q1 - 2001Q3 Finland 1980Q1 - 1989Q3* 1997Q1 - 2000Q3* France 1988Q4 - 1990Q3* 1999Q1 - 2001Q2 Germany 1989Q3 - 1990Q3* 1999Q3 - 2000Q3* Ireland 1977Q4 - 1979Q3 1987Q1 - 1990Q3* 1995Q4 - 2001Q3 Italy 1980Q4 - 1981Q3 1989Q2 - 1991Q3 1999Q1 - 2001Q2 Japan 1973Q1 - 1973Q4* 1986Q2 - 1999Q1* Netherlands 1976Q3 - 1978Q2 1988Q3 - 1990Q3* 1993Q4 - 2000Q4* New Zealand 1983Q3 - 1984Q2* 1986Q2 - 1987Q3* 1994Q2 - 1996Q4 Norway 1973Q2 - 1974Q2 1981Q2 - 1987Q3 1996Q4 - 2001Q2 Spain 1986Q2 - 1991Q2* 1998Q1 - 2001Q2 Sweden 1986Q3 - 1990Q2* 1996Q2 - 2000Q3 Switzerland 1988Q2 - 1990Q1* 1999Q1 - 2001Q1 United Kingdom 1972Q3 - 1973Q4* 1985Q4 - 1990Q1* 1999Q1 - 2000Q4 United States 1986Q1 - 1987Q3 1996Q1 - 2000Q4
* High-cost boom (annualised change at least -2.4 p.p.) ** For Australia 2004Q4 was identified as a boom quarter
Table A4: Mortgage market deregulation measures
Country Date Type of measurement
Australia 1986 Deregulation of interest rates
Belgium 1977 Partial deregulation of financial markets
Canada 1967 Deregulation of interest rates; relaxation of limits of bank borrowing
Denmark 1982 Deregulation of interest rates’ deregulation of mortgage lending
Finland 1986-1987 Deregulation of interest rates; lifting of quantitative credit controls
France 1987 Lifting of credit controls
Germany 1967 Deregulation of interest rates
Ireland 1985 Deregulation of interest rates
Italy 1988 Permanent lifting of quantitative credit controls
Japan 1994 Complete deregulation of interest rates
Netherlands 1980 Deregulation of interest rates
New Zealand 1984 Deregulation of interest rates and lifting of components of credit controls
Norway 1985 Deregulation of interest rates and lifting of credit controls (1984)
Spain 1987 Deregulation of interest rates
Sweden 1985 Deregulation of interest rates and relaxation of limits of quantitative credit controls
Switzerland 1977 Very advanced deregulation of the financial sector United Kingdom 1986 Authorization granted to building societies to extend
their activity to mortgage loans
United States 1984 Deregulation of interest rates; removal of Regulation Q
Table A5: Countries and their banking crises Country Date Australia 1989-1992 Belgium -Canada 1983-1985 Denmark 1987-1992 Finland 1991-1994 France 1994-1995 Germany 1979-1980 Ireland -Italy 1990-1995 Japan 1991-present Netherlands -New Zealand 1987-1990 Norway 1987-1993 Spain 1977-1985 Sweden 1991 Switzerland -United Kingdom 1974-1976 1984, 1991, 1995 United States 1984-1991
Source: Caprio and Klingebiel (2003)
Table A6: Type of Mortgage Rate Systems per Country
Fixed Mortgage Rate Contract Countries Variable Mortgage Rate Contract Countries
Belgium Australia Canada Finland Denmark Ireland France Norway Germany Spain Italy Sweden Japan Switzerland
Netherlands United Kingdom
New Zealand United States
