The monetary transmission mechanism and the role of house prices in Sweden : a SVAR approach

The monetary transmission mechanism and

the role of house prices in Sweden: a SVAR

approach

Hans Otto Isak Andersson

Student Number: 6371973

Masters Graduation Thesis

Economics

Faculty of Economics and Business (FEB)

2

ABSTRACT

This paper studies the role of house prices in the monetary transmission mechanism in Sweden. With the background of economic theory I estimate a five variable structural vector autoregression (SVAR) model based on that of Elbourne (2008) and Ncube et al (2011). I have used a number of restrictions to the model allowing the interest rate and house prices to react simultaneously to the information. I find that house prices react immediately to a negative interest rate shock and the decrease is strong and prolonged. On the other hand, a one percent positive impulse shock to house prices increases the consumption, the inflation and but also influences the interest rate. By assuming the reverse effect is true, then I find that the decrease in house prices seen from the interest rate shock also enhances the negative response in consumption and inflation. Thus, the role of house prices has a strong effect in the monetary policy setting. There is some uncertainty with these estimates and should thus only be seen as a guideline.

~Acknowledgments~

I would like to thank my family and friends for constantly pushing me forward.

Key Words:

House Prices, Monetary policy, Consumption, The life

cycle hypothesis, Structural Vector Autoregression, Impulse

Shock, Variance decomposition.

3

Table of content

Chapter Page ABSTRACT ... 2 ACKNOWLEDGMENTS ... 2 TABLE OF CONTENTS ... 3 LIST OF FIGURES ... 5 CHAPTER I: Introduction ... 6 1.1 Motivation ... 6

1.2 Objective of the thesis ... 7

CHAPTER II: Emperical background ... 8

2.1 Histroical developments of the Swedish housing market ... 8

2.2 Increases in house prices ... 11

2.3 Homeownership ... 12

2.4 Indebtedness ... 12

2.5 Interest rates ... 13

2.6 Consumption and housing correlation ... 14

CHAPTER III: Monetary Policy ... 16

3.1 Monetary policy objective ... 16

3.2 The role of housing in the monetary policy transmission ... 16

3.2.1 The income channel ... 17

3.2.2 The wealth effect... 18

3.2.3 The life cycle hypothesis ... 18

3.2.4 The collateral effect ... 19

3.3 Monetary policy response ... 19

4

CHAPTER IV: Methods and materials... 22

4.1 Estimation of the VAR model... 22

4.2 Defining the VAR model ... 23

4.2.1 The data ... 23

4.2.2 Unit root test ... 24

4.2.3 Lag length test ... 25

4.2.4 Johansen Cointegration test ... 25

4.3 Estimation of the VAR model... 25

4.3.1 Identification of the SVAR model ... 26

4.3.2 System of equations – Identification of restrictions ... 26

CHAPTER V: Results... 28

5.1 The interest rate shock ... 29

5.2 House price shock ... 31

5.3 Variance decompostion ... 34

CHAPTER VI: Conclusion ... 39

5

1

INTRODUCTION

The goal of this thesis is to show what has historically affected the development of house prices and what effect this has had on the overall economy and consumption. I will also describe what role house prices has in the monetary policy transmission and with this understanding use econometric methods to develop a SVAR model to empirically indentify this correlation.

1.1 Motivation

Housing plays an important role in the economic activity of a country and it is a unique asset in people’s everyday life. Housing serves two important functions: firstly housing is an investment asset and secondly it acts as a durable good that provides direct services for households. In Sweden housing wealth accounts to 54% of the average household wealth (SCB, 2011) making housing the principal asset of an individual and the largest investment an individual will undertake throughout her lifetime.

At a macroeconomic level the housing sector accounts for a considerable part of a countries welfare, wealth and gross domestic product (GDP). Therefore the housing sector influences a countries long-term development. Moreover, housing investment involves the largest borrowing requirement of a household. Because of this fact households are expected to be economically sensitive to unexpected changes in income, interest rates, equity and house prices. There is also thought to be a correlation between housing wealth and consumption. According to economic theory an increase (decrease) in house prices will make the home owner feel more (less) wealthy. This will encourage the home owner to spend more (less) of their current income today. If a change in house prices affect home owners economic behaviour, it will have a significant effect on the overall economy since private consumption is a major contributor to economic growth.

The role of the central bank is to maintain price stability in the economy. The central bank does this through its main tool, the repo rate. The repo rate directly affects banks lending rates and thus affects mortgage payments of households. This has indirect implications on household’s wealth and direct implications on income through the interest rate channel. The previous mentioned implications will affect household consumption behaviour.

6

With this significant link between monetary policy, the housing sector and the overall economy it is interesting to study the effect of how a monetary change transmits in to private consumption through house assets. In this thesis, I will provide evidence that the housing sector has an impact on consumption through the monetary policy mechanism.

1.2 Objective of thesis

I will in this thesis attempt to analyse the role of the housing market in the Swedish monetary transmission mechanism (MTM). I will aim to estimate and quantify the percentage decline in consumption expenditure in Sweden as a result of changes in housing wealth, following a monetary contraction. I use a structural vector autoregressive (SVAR) approach, based upon that of Elbourne (2008) and Ncube et al (2011) applied on the Swedish real estate price data index (all size house prices).

To get a broader understanding of the topic I give in the second chapter an overview of recent historic developments in the Swedish housing sector and its effect on the Swedish economy. The third chapter provides an overview of monetary policy and the role of house prices in the monetary policy transmission mechanism. It also contains a literature review of previous findings. The VAR and SVAR model approach is presented in chapter four together with a data description. Chapter five presents and explains the results and in chapter six concluding remarks are presented.

7

2

EMPERICAL BACKGROUND

2.1 Historic house price developments and its effect on the Swedish economy

Before the mid eighties the Swedish credit market was greatly regulated by the government. The Swedish central bank (Riksbanken) had a great influence over the banking system. However, in 1985 the Swedish government abolished interest rate regulations and deregulated borrowing requirements (OECD, 2000). Banks were now given the freedom to choose both the volume and the rate of interest to lend at. A lending boom of cheap credit started and housing consumption grew rapidly, especially among credit constrained households (Finanstidningen, 2008). The process was fuelled by rising inflation and a tax system that favoured borrowing, resulting in negative real after-tax rates. As a result, there was a rapid increase in asset prices (Jonung, 2009). Between the periods 1985 to 1991 house prices had a yearly increase of 5% to 20%. This can clearly be seen in the yearly percentage change in real estate prices (Figure 2.1), in comparison to consumer prices in Sweden. Large variability can be observed in the yearly percentage change in real estate prices, while the developments in consumer prices have been steadier.

During this time period the impact on the real economy was strong and Sweden had a high employment rate, rising consumption and low private saving ratios. Nevertheless, financial deregulation inflated home prices and increased the debt ratio as the availability of credit was much easier to obtain. This eventually led the property bubble to burst. There were different impulses that led to the instability in the Swedish economy. First, international interest rates increased, following the German reunification. Second, domestic macroeconomic policies

-15 -10 -5 0 5 10 15 20 25 30 35 40 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

Figure 2.1. Yearly percentage change of single family home prices and consumer prices in Sweden

Single family home Consumer prices

8

changed to a more restrictive fiscal policy (Englund, 1999). With increasing real interest rates, firms and household faced difficulties in paying off their mortgages. Some households began to sell of their assets to try and improve their wealth positions. However this only deepened the crisis because it created negative equity among homeowners, which spread rapidly to the banks since it left banks with large losses. The crisis spread fast with falling investment, large declines in house prices, decreasing tax revenues, increasing unemployment and the abolishment of the pegged exchange rate of the krona. In the early 1996 prices began to level out in the aftermath of the banking and property crisis (Frisell, 2010) and the Swedish economy began to recover. Low inflation and low interest rates gave rise to a more even development in house prices.

As the IT-bubble burst in the beginning of 2000, although a short damp in the increase, house prices continued to rise. This might have helped to stabilize the consumption and GDP growth in Sweden during this turbulent period. As illustrated in figure 2.2 the IT-bubble heavily affected the stock market while the price development in house prices were less volatile.

On an international level housing markets co-move and in the vast majority of OECD countries house prices have been moving up strongly since the mid 1990s. Figure 2.3 shows a comparison between real estate prices in Sweden, United Kingdom and United States. From the figure, co-movements in cross border house prices can be observed, however the amplitude of the Swedish and UK house price have been higher than the one observed in the USA. The co-movement is not a direct contagion effect from a house price shock in a

0 20 40 60 80 100 120 140 160 0 100 200 300 400 500 600 Source: SCB (2012), OECD (2012)

Figure 2.2. Share Prices, Index 2005=100 (right axis), Real estate price index for permanent houses (1981 = 100), (left axis), in Sweden

Real Estate Prices Share Prices

9

neighbouring country and the correlation is more likely to be a reflection of the relationship in macroeconomic variables between countries.

As of the recent financial crisis, Swedish house prices did not have a drastic decline in its prices, in contrast to other countries such as the United States and the United Kingdom. From figure 2.3 a temporarily dip in house prices can be seen in all three countries, nevertheless the dip in Sweden only had a stagnating effect of the positive price development and has since 2009 regained its momentum.

When the 2008/09 crisis hit, the Swedish economy had a decade of strong growth behind it, with an average yearly growth of 3.5% (European commission, 2012). However, with its dependence on exports, Swedish GDP growth fell sharply (see figure 2.4) as global trade shrank during the 2008/09 crisis and companies put investment plans on hold. Nevertheless,

-20.00 -15.00 -10.00 -5.00 .00 5.00 10.00 15.00 20.00

Figure 2.3. Percentage change in house prices from previous year

United States United Kingdom Sweden -6 -4 -2 0 2 4 6 8

Source: World data bank (2012) Figure 2.4. GDP growth (annual %)

Euro area Sweden United States

10

due to the relatively quick return of confidence to the Swedish economy, after the financial crisis, exports rebounded rapidly. Moreover, the European commission concluded that because of expansionary fiscal and monetary policies and a relatively resilient housing market, the Swedish domestic demand held up reasonably well during the crisis (European commission, 2012). This is also illustrated in the strong immediate growth in GDP, which spiked at 6% right after the financial crisis (see figure 2.4).

2.2 Increases in house prices

After the banking and property crisis in the mid ninetieths house prices have rebounded. Between 1997 and 2009, the SCB's property price index for permanent houses has shown an increase with approximately 176 percent. Deflated by the consumer price index this gives a return of 133 percent, or an average real return of 6.7 percent annually (Frisell, 2010). However apartment rents and construction costs only has increased with 13 percent and 33 percent respectively between 1995 and 2010 (Jansson et al, 2010). To some extent the upward trend in house prices is a result to changes in fundamental house price determinants. Demand side attributions such as increases in per capita income, a growing population, low inflation and tax reliefs have promoted homeownership and increased the demand for housing (Arestis, 2010). On the supply side research has pointed to the importance of planning and building restrictions, where restrictive zoning laws have kept the supply relatively low (Arestis, 2010), resulting in higher house prices. Since house prices have increased much faster than the construction costs in Sweden, the lack of new construction reflect a very high inelasticity of supply to increases in the demand for housing, which implies further increases in house prices. Nevertheless, in the short run, supply is limited due to two factors: 1) the construction industry's capacity and costs to adapt to the increasing demand and 2) the supply factors of production, particularly land (Arestis et al. 2010). In the longer term the construction industry's size should be elastic to an increase in demand, however land supply is not (at least not in urban areas). It is therefore natural that house prices react stronger to demand changes in the short run than in the long run. The Swedish real estate market is heavily characterised by a significant shortage of supply. In 2010 85 percent of the country's population lived in an urban area representing only 1.3 percent of the land (Riksbanken, 2012). As a consequence land shortage is especially prominent in the three larger metropolitan areas.

11

Over time the house price level will increase with a growing population, disposable income and land shortage (Jansson et al, 2010). On the other hand there should theoretically be a downward pressure on house prices since the cost of owning versus the cost of renting is not in balance. Because of this imbalance potential buyers should find it more advantageous to rent (Arestis et al. 2010). Swedish National Housing Credit Guarantee Board BKN (2010) have pointed out that there are tendencies for a housing bubble in the Swedish housing market since the increases cannot only be justified in terms of fundamentals. BNK (2010) concluded that the current house prices also rest to a large extent to the liberalisation of mortgage credit conditions, which has made households able to take out generous credits to low mortgage rates. In the ninetieths similar conditions occurred during the property and banking crisis.

2.3 Homeownership

The liberalization of the credit and financial markets has on the other hand made homeownership more attractive to the broader groups of income levels in Sweden. Homeownership accounts to 69,6% of all dwellings in Sweden (Eurostat, 2013). The rental housing share in Sweden is relatively low compared to the European average and accounts for 30,3% of the total housing stock. Nevertheless, there is a general assumption that homeownership in Sweden is more expensive than renting, as a consequence owning a home is a family-oriented decision rather than for financial reasons (National association of realtors, 2013).

A worry in the Swedish housing market is the growing wealth gap between those who buy a house and those who rent. Although relatively low cost of renting, households in tenancy pay on average a higher proportion of their disposable income on housing. This is because households in tenancy often have a lower disposable income (SCB, 2010) and the income growth is smaller (Wendt, 2006). If house prices remain at a high level compared to rental housing, it will become much harder to move from rented housing to owner occupied housing (Lind, 2009).

2.4 Indebtedness

The Swedish public debt has decreased significantly over the last decade and was in 2010 40% of GDP (European Commission, 2012). At the same time household debt has increased

12

and in 2010 reached a record level of 82% of GDP, 170% of disposable income (European commission, 2012). This is far higher than during the banking and property crisis in the 1990s. The bulk of household borrowing has been in the form of mortgage debt. This also reflects simultaneous increases in house prices. For this reason 90% of homeowners have a significant proportion of their savings tied to their dwellings (Riksbanken, 2010).

Household’s total assets in Sweden consist of over 71% in real assets, while the financial assets amount to approximately 28% (SCB Statistics, 2007). Price developments in the housing market therefore affect a very large proportion of household’s total wealth. With regard to the increasing levels of household debt many foreign observers, including the European Commission (2012), have expressed concern and warned of a potential bubble in the Swedish housing market. The high level of debt is a worrying factor since it has increased households vulnerability, making households more sensitive to changes in interest rates, income and asset prices fluctuations, especially in house prices itself. This is a worrying factor because high levels of debt will only enhance the negative effect on the overall economy since homeowners with negative equity are more likely to default than others (Arestis et al. 2010). Other researchers have pointed out that there is a strong lagged correlation between homeownership and unemployment (Blanchflower et al, 2013). Another possible consequence is that a higher debt burden may not only put the homeowner in negative equity but it also reduces their labour mobility since the housing market becomes less flexible. Gunnarssson (2007) showed that in areas where the proportion of homeowners increased by ten per cent could also see an increase in unemployment rate by two per cent. Although evidence suggests that the ability to move homes does affect employment, the effects do not appear to be significant (Tunstall et al, 2013). This is a topic I will not put emphasis on, nevertheless it would be an interesting topic for further research.

2.5 Interest rates

Approximately 51 percent of household mortgages in Sweden are financed with a variable interest rate (Hanson, 2009). The proportion has decreased substantially over the past two years were in 2010 almost 70 percent of all mortgage loans were to variable interest rate. Nevertheless, it is still a very high level compared to an international perspective, especially to Germany were almost all loans are to a fixed interest rate (Leonhard, 2012). The interest ratio, which measure how much disposable income goes to interest payments after tax (Finansinspektionen, 2013), is historically low due to the extremely low level of interest rates

13

in Sweden. This results in Sweden having one of the lowest mortgage costs in the western world. The large extent of variable loans also post a threat to the economy. If the repo rate would increase with over 5% within a few years, ceteris paribus, this would lead to the highest interest rate ratio since the Riksbank began its low-inflation policy (Riksbanken, 2010). As a consequence, half of all mortgage loans in Sweden are associated with a high interest rate risk.

2.6 Consumption and housing correlation

Swedish and international data show a high correlation between changes in house prices and private consumption growth (Hansson, 2009). Figure 2.6 illustrates the Swedish households' aggregate consumption expenditure in relation to developments in house prices between 1980 and 2011.

Final consumption expenditure has followed the development of house prices, indicating a correlation between the two variables. How large effect this correlation has on the economy as a whole depends on how large effect the development in asset prices has on the private consumption. In a report from Hansson (2009) showed that a 5 percent wealth effect implied a major impact on private consumption. This wealth effect contributed to private consumption growth at an average of 1.5 percent per year. Results from other empirical studies also show that increases in housing wealth is far more important than increases in financial wealth, since housing wealth is more equally spread over the population. In a study by Lettau et al. (2004), he showed that changes in financial assets have small effects on private consumption because financial wealth is more concentrated to households with higher disposable income. Because of the higher share of housing wealth across the population, increases in housing wealth are more permanent and as a consequence have greater impact on

0 100 200 300 400 500 600 0 200000 400000 600000 800000 1000000 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

Figure 2.6. Final consumption expenditure and house prices in Sweden

1980-2005 (constant prices, reference year 2000), source: SCB (2012)

Varor Riket

14

consumption. The fact that house prices barely fell during the economic turmoil of 2007-2009 is one of the reasons that the effects of the crisis was milder in Sweden than in other countries (Riksbanken 2011). This was also the case during the first years of 2000, were rising house prices during the IT-bubble prevented the economy to fall deeper than it otherwise would have done by keeping up domestic demand. This can be compared to the deep recession of the 1990's when house prices fell sharply and private consumption stagnated for several years. This correlation between house prices and consumption will be further described in the following chapter.

15

3

MONETARY POLICY

The goal of this chapter is to describe some concepts related to the basic monetary policy theory. This will give a starting point to understand the relationship between monetary policy, house prices and consumption levels considered in this thesis. A brief overview of monetary policy is given showing how a policy innovation can affect house prices and finally consumption. In last section a literature review of previous findings and studies on the topic is given.

3.1 Monetary policy objective

The objective of monetary policy in Sweden is to “maintain price stability”. The Riksbank has interpreted this objective to mean a low, stable rate of inflation (Riksbanken 2012). The underlying reason for this objective is that high inflation has negative effects on the economy, where it among other things erodes the value of money, increases economic uncertainty and redistributes wealth and income (European Central Bank, 2013). Deflation which is the opposite of inflation is neither desirable for the same mentioned reasons.

One of the key instruments the central bank can use in order to keep inflation around its two percent target, is the short term repo rate (Riksbanken, 2013). Soon after the official interest rate is changed, banks adjust their lending rates which quickly affect the interest rates that banks charge their customers for variable and fixed loans. A change in the official interest rate is almost immediately transmitted in to the economy. By raising or lowering the repo rate the central bank may indirectly affect the demand in the economy and influence inflationary pressure.

3.2 The role of housing markets in monetary policy transmission

Monetary policy works largely via its influence on aggregate demand in the economy and it affects consumption via three channels, 1) the income, 2) wealth and, 3) liquidity channels (André et al, 2010). This can also be categorised as the direct and indirect effects which is illustrated below.

16

Fig. 3.1. Simplified schematic of the monetary transmission mechanism through the housing market.

(Elbourne, 2005)

3.2.1 The income channel

The income channel works largely via the direct effect on indebted households. An increase in the interest rate increases the financial burden of households and decreases their disposable income. How fast this affects consumption depends on how fast banks change their mortgage interest rates following a monetary policy contraction and to what extent variable rates are used in the mortgage market (Elbourne, 2005). In Sweden where more than 50% of mortgage rates are to a variable interest rate the transmission effect is rapid. How the effect of an interest rate change affect total consumption in turn depends on the relative proportion of indebted and creditor households, their debts, assets and their marginal propensity to consume. Since indebted households tend to have a larger marginal propensity to consume it is likely that an increase in interest rates will have a negative effect on the consumption (Arestis et al. 2010).

17 3.2.2 The wealth effect

The wealth channel is the indirect effect of a monetary policy change. A reduction in interest rate generally stimulates the demand for housing which in turn triggers an increase in the value of real housing wealth (Elbourne, 2008). When house prices increase, the value of the house asset of the homeowner increase as well. This increase in wealth affects consumption since households will feel they are wealthier and for this reason they will feel that they have a greater ability to consume. The tendency to spend more when wealth has increased is the building block to the SVAR model in the analysis.

3.2.3 The life cycle hypothesis

The wealth effect described above is developed from the life cycle hypothesis, an economic consumption model developed in the 1950's by Franco Modigiliani. The model is build on the assumption that individuals are forward looking and plan their consumption along their known sources of income, expected future income flows and financial and real assets. Over most people’s adult lives their income will follow a reasonably predictable pattern (Carlin, et al. 2005) and thus only an unexpected change to income will make an individual change their current consumption behaviour. The model has often been used when analysing household consumption behaviour (Hansson, 2009).

In a study, made by the Federal Reserve System, Pounder (2009) demonstrated that there is a strong empirical relationship between observed consumption and expected future income. The same is for a sustained decline in interest rates and risk premiums. Lower interest rates does not only affect households choice between savings and consumption, but also has a direct effect on the value of both real and financial assets through the wealth effect of a monetary policy change (OECD, 2000).

Gourinchas et al. (2002) demonstrated the same conclusion were they also estimated that households discount the future at modest rates and are not particularly risk averse. However, their results also implied that the older you get the more households save actively for their own retirement purposes.

18 3.2.4 Collateral effect

Changes in house prices have significant effects on consumption, not only through the wealth effect but also through the collateral effect. For the life cycle hypothesis to properly function, the condition of a well working credit market must exist in order for individuals to borrow against their future expected lifetime (Ncube et al 2011). In an uncertain world with imperfect financial markets and asymmetric information, credit constrained consumers can use their housing collateral when borrowing, in order to smooth out their consumption. An increase in the value of housing assets will allow more borrowing against the asset in order to finance current consumption, this outcome is called the collateral effect (Arestis et al. 2010). Through collateral the information problem is reduced because good collateral significantly decreases losses to the lender if the borrower defaults on the loan. It also reduces the incentives for the borrower to take on excessive risk because the borrower now has something to lose (Mishkin, 2007). This has been the case in Sweden were increases in house prices has meant that households have been able to borrow more money against the real value of their houses (Hansson, 2009).

3.3 Monetary policy response

Since house prices affect the overall economy, the task of managing the economy in an effective way requires that monetary policy authorities respond to changes in house prices (Mishkin, 2007). However, the question is not whether the central bank should respond to house prices movements, but “how”. Since a few years the Swedish Riksbank uses a macroeconomic general equilibrium model called the Swedish Riksanks Aggregate Macro Model for Studies of the Swedish Economy (RAMSES). This model is used to make predictions, interpret economic developments and calculate effects of different monetary policy assumptions (Adolfson, 2007). According to this economic model, it may have major negative consequences for the real economy if monetary policy would try to prevent rising house prices. In the RAMSES model monetary policy only have a relatively minor effect on house price developments. It therefore requires large interest rate hikes to dampen rising house prices, which would have a negative effect on household consumption and business investment leading to a lower GDP development (Jansson et al, 2010).

On the other hand economists such as Roubini (2006) have argued in favour of monetary targeting and that a moderate interest rate response can have an impact on house price

19

bubbles and reduce the economic distortions caused by them. Roubini argued that the argument of trying to burst a house price bubble would require a large interest response which could cause a severe recession, found to be incorrect both in theory and practice. However he has suggested that monetary policy should only respond to asset bubbles in a cautious and moderate matter. In contrast Minshkin (2007) argued that there are three key assumptions needed in order to prevent bubbles with monetary policy targeting. The central bank can only target a bubble if they first can, identify a bubble in progress. Secondly, the central bank cannot appropriately deal with the consequences of a burst bubble, therefore pre-emptive actions are needed to prevent a bubble. The third assumption needed in order to justify a special focus on house prices, in the conduct of monetary policy, is that a central bank knows the monetary policy needed to deflate a bubble. However, if the central bank knows that there is a bubble and if they don’t have any informational advantage then the market will know too and the bubble will burst. The effect of an interest rate change on asset-price bubbles is then highly uncertain. Although some theoretical models suggest that raising interest rates can diminish the acceleration of asset prices, others suggest that raising interest rates may only cause a bubble to burst more severely, doing even more damage to the economy (Minshki, 2007). The central bank should on the other hand respond to the bursting bubbles in a timely manner to keep the effect after a burst at a manageable level.

3.4 Literature review and previous findings

In this section a literature review of previous findings and studies on the topic will be given. Many economists have been studying the role of house prices and the effect thereof, because of house prices important link with the economy.

Since the 2007 financial crisis, the responsiveness of house prices to monetary policy have been studied quite extensively. Nevertheless, this topic was also an issue before the outburst of the crisis. Elbourne (2005) was one of the first to use a SVAR model in order to examine the effect the UK housing market had on consumption through the transmission of monetary policy. Elbourne measured the effect of a 100 basis point shock to short term domestic interest rates. He found that the response was in line with the predictions of economic theory where consumption, prices and the demand for money decreased. In his model retail sales decreased by 0.4% and house prices fell by 0.75% following a contraction in monetary policy. Moreover the effect of a positive house price shock increased consumption by 0,07%

20

in his model. By combing these central estimates Elbourne concluded that there was evidence that housing wealth did play a role and that 15% of the fall in consumption after an interest rate shock was due to changes in house prices. Ncube et al (2011) used a similar SVAR model based on the Elbourne (2008) model applied on South African disaggregated house price data. In their thesis they showed the importance of the interest rate effects working through both housing wealth and the credit channel in influencing real spending. They found that both consumption and house prices fell in response to a 0.5% interest rate increase. However, when assessing the direct effect of a interest rate shock on consumption compared to the indirect effect they concluded that the direct effects of high interest rates on consumption are more important than the indirect effects. Therefore Ncube et al (2011) concluded that a monetary policy tightening can only marginally weaken inflationary pressure arising from excessive consumption operating through housing wealth.

Similar results were showed in Hong Kong Monetary Authority’s (2008) VAR analysis of the transmission mechanism of interest rate movements to output growth and inflation in Hong Kong. In their model a 100 basis point increase in the three month HIBOR led to across the board declines in real GDP, property prices and headline inflation. However the impact of interest rate shocks through the property price channel on real GDP growth was relatively small. Studies have also been done on Swedish house price and consumption data. Chen (2005) conducted a VECM co-integration model with a permanent-transitory variance decomposition framework on quarterly Swedish data between the period 1990 to 2004. Chens key findings was that there is a strong correlation between aggregate consumption, disposable income, housing and financial wealth. He found that the long run elasticity of total consumption with respect to net housing was 0.11. In the short run, total consumption is expected to increase by 0.064 percentage point following a 1% increase in housing wealth.

In support to Chens findings (2005) Björnland Et Al (2010) showed that monetary policy has a strong and prolonged effect on house prices, emphasizing the role of house prices in the monetary policy transmission mechanism. They analysed the role of house prices in the monetary policy transmission mechanism in Norway, Sweden and the UK, using a structural VAR approach. Following a contractionary interest rate shock, house prices did not only fall immediately, but also had a lasting effect were house prices fell up to 1,5-2,5 years after the shock. Nevertheless, the strength and timing of the response varied from one country to another, indicating that housing may play different roles in the monetary policy setting

21

between countries. One reason for this is that there are large significant differences in the mortgage product offerings in developed countries were countries differ in terms of the market share of adjustable versus fixed rate mortgages (Lea M. 2010). Liberalised mortgage markets with variable loans are important for the monetary transmission mechanism to transmit in to the economy. Kiss and Vadas (2007) demonstrated that due to long non-callable loans in the Hungarian mortgage market, there was a delay between key policy interest rate and the mortgage market. This situation caused a limiting effect of the interest rate on consumption through changes in house prices. In Giuliodori (2002) research of the monetary policy transmission, he found that where financial markets are more liberalised house prices play an important role in the transmission of interest rate shock to household consumer spending. Walentin (2011) studied the effect of how the monetary transmission mechanism would be affected by a change in the loan to value ratio in Sweden, or more broadly household indebtedness. The main results from his research showed that a large proportion of the population was collateral constrained. This implies that house prices have a substantially effect on macroeconomic variables through the collateral effect, housing used as collateral for loans then consequently reinforced the effects of monetary policy. The monetary transmission mechanism then works better if the loan to value ratio is higher together with a liberalised mortgage market. As a final observation Yang (et al. 2010) did a study on the heterogeneous effects of monetary policy on regional house prices in Sweden. They used a multivariate persistent shock metric to examine the impact of a change of the short-run interest rate on the property market in Sweden. In their study they found that there are large regional differences to the effect of a monetary policy change. The impact of an interest rate change is highly sensitive in the most populous areas of Sweden and for this reason a monetary policy contraction will have different outcomes on house prices in different regions.

22

4

METHODS AND MATERIALS

The empirical method adopted in this paper is based upon that of Elbourne (2008) and Ncube (2011) and follows a vast majority of empirical research and literature on the topic.

4.1 Estimation of the VAR model

The first step in developing the SVAR model is to estimate a VAR model. The estimation of a VAR model is straightforward and each equation is estimated as an OLS regression on the explanatory variables.

I assume the economy has the structural form equation [1] as in Ncube (2011) and Elbourne (2008).

[1]

A(L)x

t

= ε

t

A(L) is the matrix polynomial in the lag operator L and t x is an nx1 vector of explanatory variables. ε1 is nx1 structural disturbance vector which are serially uncorrelated. The residual

vector variance is var(

ε

t) = Ω with diagonals as variances of structural disturbances. These

structural disturbances are assumed to be mutually uncorrelated. We then estimate the reduced form equation (VAR)

[2]

x

t

= B(L)x

t

+

ut

where B(L) is a matrix polynomial with a constant term in lag operator L and var(ut)= Σ= E(ut

u′t). I adopt the structural VAR modelling in which non recursive structures are allowed while

still giving restrictions only on contemporaneous structural parameters.

Letting A0 to denote the contemporaneous coefficient matrix in the structural form of

non-singular matrix at lag zero in A(L) and letting A+ (L) be coefficient matrix in A(L) at strictly positive lags excluding the contemporaneous coefficient A0, we express A(L) as

23

The parameters in the structural form equation and those in the reduced form are related by

[5] B(L) = −A0−1A+ (L)

The structural disturbance and the reduced form residuals are related by εt = A0ut which

implies that Σ = A0−1ΩA0−1.

4.2

Defining the VAR model

4.2.1 The data

I define a small open economy (Sweden) for the estimate of the VAR, with 5 endogenous variables. The variables used in the model are the Swedish consumer price index (cpi, 1980 = 100), final consumption expenditure by households as a proxy for consumption, the three month domestic interest rate, Swedish house price index and the exchange rate of Swedish kronor (SEK) to Dollar ($). The money-market interest rate represents the monetary policy stance of the Swedish central bank (Riksbanken) which reflects the main instrument in the monetary policy setting. All data are from the Swedish Riksbank. The house price index is calculated from data of owner occupied one- and two- dwelling buildings, representing the whole country. All variables are in logarithms except the quarterly money-market interest rate which is estimated in levels.

The plot of variables below (figure 4.1) shows the time trend of the five variables used in the model. House prices index, consumption and CPI all show a step upward trend while the exchange rate (SEK/$) and the money market interest face volatility over the period.

24

Figure 4.1 Plot of variables

4.2.2 Unit root test

Following standard practice before estimating the model I check that all variables are stationary to avoid regression problems associated with unit roots. The Augmented Dickey Fuller (ADF) unit root test support the null hypothesis that all variables have a unit root. This means that in level the variables do not evolve around a zero mean. In order to transform the variables to stationary variables I take the first difference and all variables except the interest variable (i) becomes stationary.

240000 280000 320000 360000 400000 440000 94 96 98 00 02 04 06 08 10 Consumption 240 250 260 270 280 290 300 310 320 94 96 98 00 02 04 06 08 10

Consumer Price Index

100 200 300 400 500 600 94 96 98 00 02 04 06 08 10

House Price Index

5 6 7 8 9 10 11 94 96 98 00 02 04 06 08 10

SEK/$ Exchange Rate

0 2 4 6 8 10 94 96 98 00 02 04 06 08 10 Interest Rate

25 4.2.3 Lag length

There are various methods to choose for the lag length for the VAR model. In the following step the appropriate lag length has been selected for the model using the lag selection mechanism based on Akaike Information Criterion (AIC), Schwarz Information Criterion (SC), Hannan and Quinn Information Criterion (HQ) and Final Prediction Error Information Criterion (FPE). With five different criteria for lag length selection, there is the possibility of disagreement. AIC, FPE and LR suggested using 4 lags in the VAR model, while SC and HQ points out 2 lags for the optimal model. Based on the majority rule I choose four lags. Longer lag lengths do not change the main qualitative results, although the loss of degrees of freedom could affects the statistical significance of the impulse response functions (Giuliodori, 2006).

4.2.4 Johansen Cointegration

To test the co-integration relationship in the model to see if the variables have a long run association shape or not I apply the Johansen integration test. From the Johansen Co-integration Trace test it was indicated that there are 5 co-Co-integrations between the 5 variables in the model at a confidence level of 95%. In multivariate models it is possible for several series that are non-stationary with unit roots to have a linear relationship that produces stationary disturbances and it is possible to remove unit roots without taking differences (Shields, 2007). If co-integration among the variables exists, the system’s dynamics can be consistently estimated in levels in a VAR model (Pedram et al, 2011). An alternative approach is to estimate the VAR model in first differences. This solution, however, implies discarding the information contained in the levels, and could lead to misspecification and over-differentiation (Giuliodori, 2006).

4.3 Estimating the VAR

I estimate two VAR models, one unrestricted and one restricted. In the restricted VAR model I include two dummy variables in order to account for exceptionally high interest rate in Q2 1995 and Q3 2001. However, the likelihood ratio statistics rejected the restricted VAR model including the dummy variables in favour for the unrestricted VAR model, thus no dummy variables are included. Each VAR equation contains a large number of coefficients that are individually insignificant. This is common in a VAR model and is symptomatic of the over-parameterisation which often is required by the VAR specification (Shields, 2007). The

26

individual estimates might not be BLUE in the model, however since I will interpret the VAR model by looking at the effect of an unanticipated shock, standard regression-type interpretations are not used. I do not individually interpret the coefficients and I persist with the standard VAR estimation. The most common way of analysing VAR models are through impulse response functions and variance decompositions.

4.3.1 Identification of the SVAR model

SVAR models have become a popular tool in the analysis of the monetary transmission mechanism and sources of business cycle fluctuations. SVAR models are not well suited for policy simulations, however it is a useful tool to analyse the dynamics of a model by subjecting it to an unexpected shock (Gottschalk, 2001), for this reason it has an advantage in the analysis of the monetary transmission mechanism.

4.3.2 System of equations – identification of restrictions

With a five variable VAR model I identify five structural shocks, with the interest rate shock (Ɛi) and the house price shock (ƐHP) as of primary interest. I loosely identify the other three shocks as an inflation shock (ƐCPI), consumption shock (Ɛcons) and exchange rate shock (Ɛ$/sek).

[6] – Short Run Restrictions

Each variable in the model does either react to another variable (represented by “a”), or being restricted with a zero contemporaneous restriction (represented by “0”). The matrix thus consists of 10 contemporaneous restrictions. In the baseline model (6) ucons, uCPI, uPH, uSEK / $

and uiare residuals from the reduced form equations, while

ε

cons,

ε

CPI,

ε

HP,

ε

ER and

ε

i are

27

Consumption expenditure by households (ucons) is allowed to react immediately to shocks to

its own variable. I assume the CPI (

ε

_CPI) depends on changes in consumption, as an increase in this variable will lead to higher inflation. The model follows standard restrictions in a closed economy by placing consumption and inflation above the interest rate in the ordering

of the matrix as described in (6). The house price variable (

ε

HP) reacts contemporaneously to

the consumption expenditure by households (ucons) and the interest rate (ui).

The exchange rate is assumed to be a financial variable which reacts quick to all information. For this reason I let the exchange rate react to all variables except real house prices (Ncube et al, 2011). The Riksbank conducts a flexible inflation targeting (Riksbanken, 2013) and thus the interest rate depends on all variables.

28

5.

RESULTS

I estimate the SVAR model using four lags suggested by the lag length criteria and an intercept in the estimation.

Through the SVAR model I attempted to quantify the impact of a contractionary interest rate shock on private consumption through the house price channel. I will interpret the SVAR model by looking at the effect of an unanticipated interest rate and house price impulse shock to the model. I am applying the most common way of analysing a VAR model by implying an impulse response function and variance decompositions.

The impulse response function is based on residuals of each variable and tracks the impact of any variables on others in the system. It is an essential tool in empirical causal analysis and policy effectiveness analysis (Lin, 2006). Today there are several different methods to estimate the contemporaneous effects between the residuals. The standard approach have been to implement the Cholesky decomposition, however the problem using this approach is that it may give different impulse response functions depending on which order the variables are entered in to the matrix (Shields, 2007). I will for this reason attempt to use a more recent method referred to as a “generalized impulse response function” (GIRF). A generalized impulse response provides a tool for describing the dynamics in a time series model by mapping out the reaction, for example, in house prices to a one standard deviation shock to the residual in the interest rate equation (Warne, 2008).

In order to have an ordering invariant estimation Pesaran and Shin (1998) defined the following impulse response function.

[7] – The generalized impulse response function (GIRF)

GIRFx(n, δj, Ωt-1)= E(Xt+n|μjt = δj, Ωt-1) – E(Xt+n| Ωt-1)

The generalized impulse response function (GIRF) is made up by the known history of the economy up to time t- 1by the non-decreasing information set Ωt-1 in function xt at horizon n.

Central to the generalized impulse response function is δj, the hypothesized vector of shocks

(Pesaran, 1997). The GIRF follows the idea of nonlinear impulse response function and computes the mean impulse response function.

29

When one variable is shocked, other variables also vary as is implied by the covariance. GIRF computes the mean by integrating out all other shocks making it unaffected by ordering or variables (Lin, 2006). The choice of δ is therefore central to determining the time profile for any generalized impulse response function.

5.1 The interest rate shock

I first examine the impulse responses due to an interest rate shock. Figure 5.1 shows the impulse response functions of an interest rate shock (one standard deviation) of a contractionary monetary policy (100 basis point increase in the interest rate) on the range of variables affecting the Swedish housing market. The dotted lines shows 95% bootstrapped confidence intervals calculated using Monte Carlos method with 1000 replications. The responses are in line with the predictions of economic theory, where consumption and house prices fall.

30

Figure 5.1 Interest Rate Shock

From the impulse response analysis of an interest rate shock to house prices there is indeed a link between these two variables. The response of decline in house prices is immediate. The

-.010 -.008 -.006 -.004 -.002 .000 .002 .004 1 2 3 4 5 6 7 8 9 10

Response to a Monetary Policy Shock. Consumption -.002 -.001 .000 .001 .002 .003 .004 1 2 3 4 5 6 7 8 9 10

Response to a Monetary Policy Shock. Inflation -.012 -.008 -.004 .000 .004 1 2 3 4 5 6 7 8 9 10

Response to a Monetary Policy Shock. House Price Index

-.06 -.04 -.02 .00 .02 1 2 3 4 5 6 7 8 9 10

Response to a Monetary Policy Shock. SEK/$ Exchange Rate

-.4 -.2 .0 .2 .4 .6 .8 1 2 3 4 5 6 7 8 9 10

Response to a Monetary Policy Shock. Interest Rate

Response to Generalized One S.D. Innovations ± 2 S.E. The vertical axis shows the (%) change over the time horizon.

31

decline is however weak and falls over the curve were it reaches its lowest level three quarters after the contraction at minus 0,006%. The probability bands are widening the further out on the curve, emphasizing the uncertainty in the response. The possible effect of an interest rate shock to house prices is an immediate increase in the user costs of housing, due to higher mortgage rate payments. This in turn decreases the demand for housing and hence decreases the price of housing itself. House prices gradually recover after the shock and move back towards the baseline in quarter eight. The outcome is in line with what Elbourne (2008) found in the UK housing market. Bjornland et al (2010) demonstrated similar results in their SVAR model on Swedish data, however their results showed an stronger effect of interest rates on house prices. Nevertheless, both papers showed that an increase in the interest rate had a negative effect on house prices.

From the impulse shock there is a negative association between an increase in the interest rate and consumption expenditure. When the interest rate increases, current consumption becomes relatively more expensive and households will tend to substitute away from current consumption towards savings. The decrease in consumption is almost immediate to an interest rate shock. The impact of the shock is having a lasting effect on the consumption and decreases to minus 0,005% in quarter nine. The extent of the decrease out at the curve is somewhat uncertain since the probability bands are widening.

The effect of an interest rate shock on inflation shows that the rate of inflation is slowly decreasing over the curve. The slow effect of an increase in the interest rate could be explained by sticky prices and that the response by the central bank might have been too weak to curb the inflation.

5.2 House Price Shock

In order to understand the role played by house prices in the transmission mechanism, I investigate the reverse causation. I look at the impulse response in the price level and the consumption variable to a one off shock in house prices and then see how much house prices fall, following an interest rate shock.

By combining the response of consumption to a house price shock and the response in house prices to an interest rate shock, can give an idea of the importance of house prices in the

32

transmission mechanism. The idea is to see how big effect an interest rate shock has to consumption through the effect it has on a change in house prices.

Following a one standard deviation shock in house prices (1% positive shock to house prices) there is an observed correlation between the increase in house prices and consumption. The increase is immediate and increase with 0.002% in quarter two. The observed correlation could be an effect through either the life-cycle wealth effects or through the collateral effects, whereby house price rises relax borrowing constraints. After its peak in quarter three consumption moves back towards the baseline.

Inflation is initially negative and increases first in quarter two. Inflation reaches its highest point at 0.001% in quarter five. The increase in inflation, although not certain, could be a lagged reaction to the increases in spending through the wealth or collateral effects.

Following a shock in house prices, interest rates increases with 9 basis points five quarters after the shock, which is in line with the estimates of Bjornland et al (2010). The response of the interest rate is slow. One of the reasons for this is that monetary policy responds to asset prices only over time, if they are seen to diverge from the levels which the central bank feels comfortable with (Assenmacher-Wesche et al, 2008). Thus, a onetime shock to house prices might not trigger an immediate increase in the interest rate. Nevertheless, house prices have risen hugely in Sweden, this response might not be seen to be large, however, over time the increase is extensive. By comparing the response of the interest rate to the inflation rate there is a possibility that the monetary policy responds to the increase in the inflation and not to house prices itself.

33 Figure 5.2 House Price Shock

-.006 -.004 -.002 .000 .002 .004 .006 1 2 3 4 5 6 7 8 9 10

Response to a House Price Shock. Consumption -.002 -.001 .000 .001 .002 .003 1 2 3 4 5 6 7 8 9 10

Response to a House Price Shock. Inflation -.010 -.005 .000 .005 .010 .015 1 2 3 4 5 6 7 8 9 10

Response to a House Price Shock. House Price Index

-.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 .05 1 2 3 4 5 6 7 8 9 10

Response to a House Price Shock. SEK/$ Exchange Rate

-.3 -.2 -.1 .0 .1 .2 .3 .4 .5 1 2 3 4 5 6 7 8 9 10

Response to a House Price Shock. Interest Rate

Response to Generalized One S.D. Innovations ± 2 S.E. The vertical axis shows the (%) change over the time horizon.

34 5.3 Variance decomposition

Variance decomposition provides a different method of depicting the system dynamics. While the impulse response function traces the effect of a shock to an endogenous variable on the other variables in the VAR, variance decomposition decomposes the variation in an endogenous variable into the component shocks to the endogenous variables in the VAR. The variance decomposition gives information about the relative importance of each random innovation to the variables in the VAR (Hall, 2004).

[8] – Variance decomposition

To find the impulse response function of, say, xt to ε1,t, we would set ε1,t = 1, ε2,t = 0, and all

subsequent shocks are zero in expectation. The impulse response on impact would be C1,1(0),

the response after two periods would be C1,1(1), and so on. We could do the same for variable

2. Our generic definition would be that Ci,j (h) is the impulse response of variable i to shock j at horizon h. The matrix B governs the impulse responses of the variables to the shock on impact – for this reason it is sometimes called the impact matrix or the matrix of impact multipliers. The forecast error of a variable at time t is the change in the variable that couldn’t have been forecast between t − 1 and t. This is due to the realization of the structural shocks in the system, t . We can compute the forecast error over many different horizons, h (Sims, 2011). The forecast error variance at horizon h = 0 for each variable is:

E

t

x

t

– E

t-1

x

t

= C

1, 1

(0)

Ɛ1,t

+ C

1, 2

(0)

Ɛ2,t

E

t

z

t

– E

t-1

z

t

= C

2, 1

(0)

Ɛ1,t

+ C

2, 2

(0)

Ɛ2,t

The forecast error variances are just the squares of the forecast errors (since the mean forecast error is zero). Let Ωi(h) denote the forecast error variance of variable i at horizon h. Then at h

= 0, this is simply:

35

Ω

2

(0) = C

3,1

(0)

2

+ C

3,2

(0)

2

The above follows from the assumptions that the shocks have unit variance and are uncorrelated. The forecast error of the variables at horizons h = 1 is:

E

t

x

t+1

− E

t−1

x

t+1

= C

1,1

(0)

Ɛ1,t+1

+ C

1,2

(0)

Ɛ 2,t+1

+ C

1,1

(1)

Ɛ1,t

+ C

1,2

(1)

Ɛ2,t

E

t

z

t+1

− E

t−1

z

t+1

= C

2,1

(0)

Ɛ1,t+1

+ C

2,2

(0)

Ɛ 2,t+1

+ C

2,1

(1)

Ɛ1,t

+ C

2,2

(1)

Ɛ2,t

The forecast error variances are then:

Ω

1

(1) = C

1,1

(0)

2

+ C

1,2

(0)

2

+ C

1,1

(1)

2

+ C

1,2

(1)

2

Ω

1

(1) = C

3,1

(0)

2

+ C

3,2

(0)

2

+ C

1,1

(1)

2

+ C

1,2

(1)

2

To go to more periods, we can then define the forecast error variances recursively as follows:

Ω

i

(0) = C

i,1

(0)

2

+ C

i,2

(0)

2

Ω

i

(1) = C

i,1

(1)

2

+ C

i,2

(1)

2

+ Ω

i

(0)

Ω

i

(h) = C

i,1

(h)

2

+ C

i,2

(h)

2

+ Ω

i

(h − 1)

More generally, for a n variable system, the total forecast error variance of variable i at horizon h in a n variable system is:

A forecast error variance decomposition – or just variance decomposition for short – is a way to quantify how important each shock is in explaining the variation in each of the variables in the system. It is equal to the fraction of the forecast error variance of each variable due to each shock at each horizon. Let ωi,j (h) be the forecast error variance of variable i due to shock j at horizon h (Sims, 2011). This is:

36

The fraction of the forecast error variance of variable i due to shock j at horizon h, denoted φi,j (h), is then the above divided by the total forecast error variance:

5.3.1 Variance decomposition of SVAR model

Figure 5.3.1 Variance decomposition of house prices

The unforeseen variation in house prices are largely self-determined. Nevertheless the further out on the curve other variables grow in importance, in quarter six interest rates (9,48%) and inflation (25,1%) make up the largest variation in house prices after house prices itself. The interest rate is of importance to the house price variation. This was also seen from the monetary policy innovation shock, where after the shock house prices fell. If the reverse causation is true, where house price innovations have a considerable bearing effect on

0 20 40 60 80 100 1 2 3 4 5 6 7 8 9 10

Percent of house price v ariance due to GDP

0 20 40 60 80 100 1 2 3 4 5 6 7 8 9 10

Percent of house price v ariance due to Inf lation

0 20 40 60 80 100 1 2 3 4 5 6 7 8 9 10

Percent of house price v ariance due to house prices

0 20 40 60 80 100 1 2 3 4 5 6 7 8 9 10

Percent of house price v ariance due to $/SEK exchange rate

0 20 40 60 80 100 1 2 3 4 5 6 7 8 9 10

Percent of house price v ariance due to interest rate

Variance Decomposition of house prices

37

fluctuations in consumption and interest rates, it will point to a significant role of house prices in the monetary policy setting.

Initially the consumption variable is also the most important cause of the variation of its own variable. However in quarter ten inflation and interest rate become of larger importance, accounting to almost half of the variation in consumption. Nevertheless house prices do play a role, although not to the same extent as the previous mentioned variables. In quarter ten it contributes to almost 5,5% of the unforeseen variation in consumption, thus makes it an important driving factor. This also goes in line with the impulse response reaction, where consumption increased with a house price shock.

Figure 5.3.2 Variance decomposition of interest rate

The variance decomposition of an interest rate shock gives an idea of the variables in the VAR model, that Riksbanken reacts to when setting the interest rate. The largest proportion of the variance in the interest rate (after interest rate itself) is inflation, accounting to 31,7% already in quarter one. Consumption expenditure make up to between 2,2% up to 20% of the unforeseen variation in the interest rate during the period, while house prices only make up

0 10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10

Percent of the interest rate v ariance due to GDP

0 10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10

Percent of the interest rate v ariance due to CPI

0 10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10

Percent of the interest rate v ariance due to house prices

0 10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10

Percent of the interest rate v ariance due to the exchange rate

0 10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10

Percent of the interest rate v ariance due to the interes rate

Variance Decom pos ition of the interes t rate

38

approximately 2% over the period. This results into making the contribution to the interest rate from house prices relatively insignificant. This reflects the monetary policy stand of the Riksbank, where they take the risk of sharp corrections of house prices in the future into their risk assessment. However, the Riksbank does not fully react to these changes in house prices, since this would need large rate hikes.

The contribution from monetary policy to the growth in house prices is small and house prices growth fall immediately after a monetary policy shock. Nevertheless the reverse, where a monetary policy reacts to a house price shock, is less extensive and interest rates only marginally increase. Moreover, looking at the variance decomposition of the interest rate house prices only play a small role in the interest rate setting. From this it can be concluded that monetary policy has a great influence on house prices, however the central bank does not fully react to a one off shock in house prices. The relative importance of house prices shocks is non trivial and can be seen in the variance of consumption growth. Following a house price shock consumption growth increases immediately and the contribution from an increase in house prices is up to 5,5%. The contribution from a growth in house prices to inflation is also large and accounts up to 6,6% over the period.

The transmission of a change, in the interest rate to consumption through house prices, is visible. A decrease in the interest rate has an immediate positive effect on house prices. This increase in house price growth is immediately transmitted in to the growth of consumption. Therefore house prices play a significant role in the systematic part of monetary policy.

However the rate of estimation of the indirect effect in my thesis, is less than the prediction. In my estimations a 1% real house price shock raises consumption by 0,012%. Following a monetary policy shock house prices fall only by 0,0066%, showing strong resistance to monetary policy. A combination of the central estimates of the responses shows only a marginal fall in consumption, due to the change in house prices caused by a change in monetary policy. Nevertheless, the direct effects are strong. Therefore it is of importance not to overlook the effect monetary policy has on house prices and consequently on consumption.