Master’s Thesis - May 15th, 2017

Thomas Fiskerstrand Sondre Riise Kjelstrup

Are we currently experiencing a bubble in the Norwegian housing market?

MSc. Finance & Investments Copenhagen Business School Supervisor: Jørgen Bo Andersen

Number of characters: 199,018 / Number of pages: 120 Abstract The Norwegian housing market has experienced an extreme growth in prices between 1990-2017. With this growth in mind, and the great amount of attention the topic receives from both the media and the general public, the Norwegian housing market has become a favorite among experts with regards to predicting and explaining current and future developments. Our goal with this master’s thesis is to investigate whether the recent price development can be supported by fundamental factors. Simply stated, are we currently experiencing a bubble in the Norwegian housing market?

For our thesis, we have chosen to use previously made housing price models to conduct a comparative empirical analysis of the housing price development. The models we have chosen to use are the Hodrick- Prescott Filter (HP filter), the Price-to-Rent ratio and Tobin’s Q. For all these models we have also compared our results with Denmark and Sweden to be able to better explain if the development in Norway is abnormal. We also re-estimated the widely-used Jacobsen & Naug model to investigate if it still explains the development in prices and how the many variables have changed in terms of how they affect housing prices. All these housing price models include specific fundamental factors both on the supply and demand side of housing. The magnitude of these fundamental factors are further analyzed to investigate the underlying factors of the housing price development.

Interestingly enough, all housing price models except for the HP filter applied on Norway, point to bubble tendencies in the Norwegian housing market. This is also concluded from Case & Schiller’s seven criteria for a housing bubble at the end of our thesis. However, when investigating the fundamental factors on both the supply and demand side of the housing market the price growth is supported. This is mostly led by a record low key rate, the large deviation between supply and demand in housing, a low and stable unemployment rate, steadily increased disposable income, population growth and positive expectations about future housing prices. With that said, while housing prices have increased, the debt burden among the general public has also continued to increase, making politicians extremely weary of the situation. This is shown by the recent regulations the government has put on the banks’ lending policies. In conclusion, our investigation for this master’s thesis leads us to believe that there is currently no bubble in the Norwegian housing market. Table of Contents

1. Introduction...... 5 1.1. Motivation ...... 5 1.2. Problem Statement ...... 6 1.3. Methodology ...... 7 1.4. Delimitation ...... 7 1.5. Data ...... 8 1.6. Thesis Structure ...... 9

2. Historical Development of the Norwegian Housing Market...... 10

3. Bubble Theory ...... 14 3.1. Case & Shiller’s Market Characteristics ...... 15 3.2. Euphoric and Non-Euphoric Bubbles ...... 16

4. Supply and Demand in the Housing Market ...... 18 4.1. Supply and Demand Theory ...... 18 4.2. Supply and Demand in the Housing Market ...... 18 4.2.1. Supply ...... 19 4.2.2. Demand ...... 22

5. Comparative Empirical Analysis ...... 27 5.1. Hodrick-Prescott Filter ...... 27 5.1.1. Limitations ...... 28 5.1.2. Empirical Testing ...... 29 5.1.3. Comparison ...... 32 5.2. Price-to-Rent Ratio ...... 37 5.2.1. Assumptions ...... 41 5.2.2. Data Material ...... 42 5.2.3. Empirical Testing ...... 43 5.2.4. Fundamental vs. Real P/R Ratio ...... 45 5.2.5. Limitations ...... 47 5.2.6. Comparison ...... 47 5.2.7. Data Material – Sweden ...... 48

2 5.2.8. Empirical Testing – Sweden ...... 49 5.3. Tobin’s Q ...... 52 5.3.1. Marginal and Average Q ...... 53 5.3.2. Tobin’s Q and the Housing Market ...... 54 5.3.3. Limitations ...... 55 5.3.4. Data ...... 56 5.3.5. Empirical Testing ...... 57 5.3.6. Comparison - Denmark ...... 59 5.4. Conclusion ...... 61

6. House Price Models ...... 62 6.1. MODAG ...... 62 6.2. Jacobsen and Naug ...... 63 6.2.1. Weaknesses and Discussion ...... 66

7. Fundamental Analysis of Supply Side ...... 68 7.1. New builds ...... 68 7.2. Building costs ...... 71 7.3. Cost of land ...... 73 7.4. Bank Regulations ...... 75 7.4.1. Collateral in real estate ...... 75 7.4.2. Loan-to-value ratio ...... 75 7.4.3. Interest rate increases and Interest-only mortgages ...... 76 7.5. Conclusion ...... 77

8. Fundamental Analysis of Demand Side ...... 78 8.1. Disposable income ...... 78 8.2. Unemployment ...... 80 8.3. Interest Rate ...... 81 8.4. Population growth ...... 83 8.5. Demographics ...... 84 8.6. Housing Taxation ...... 86 8.6.1. Tax on Housing Capital ...... 86 8.6.2. Property Taxes ...... 87

3 8.6.3. Tax Deductions on Interest Expenses ...... 87 8.6.4. Tax on Sale Profit ...... 87 8.6.5. Tax on Rental Income ...... 88 8.7. Expectations ...... 88 8.8. Conclusion ...... 90

9. Correlation Analysis ...... 91 9.1. Analysis ...... 91

10. Re-estimation of the Jacobsen and Naug model ...... 94 10.1. Expectation Variable ...... 94 10.2. Re-estimation of Jacobsen and Naug model ...... 98 10.3. Testing the model ...... 100 10.3.1. Testing for Stationarity ...... 101 10.3.2. Testing for Autocorrelation ...... 104 10.4. Interpretation of the Coefficients ...... 106 10.5. Discussion of model ...... 108 10.6. Conclusion ...... 111

11. Case and Schiller’s Seven Criteria for a Housing Bubble ...... 112 11.1. See housing as an investment ...... 112 11.2. Widespread agreement of an increase in prices ...... 113 11.3. Exaggerated expectations, excitement and word of mouth ...... 113 11.4. Sense of urgency in buying a home ...... 113 11.5. Simple or simplistic theories ...... 114 11.6. The occurrences of sales above asking prices ...... 114 11.7. Perception of risk ...... 115 11.8. Conclusion ...... 116

12. What do the Experts and Professionals Say? ...... 117

13. Conclusion ...... 119

14. References ...... 121

15. Appendix A ...... 133

4 1. Introduction Since the beginning of the 1990’s, housing prices in Norway have increased by 500% (Holberggrafene, 2017). In this 25-year timespan we have experienced a continuous increase in the housing market, only stagnated by minor corrections. Last time we saw a similar growth pattern was in the early 80’s, which resulted in a major crash. With this in mind many experts and professionals are debating whether the growth we are experiencing now can backed up by fundamental factors, or if we are experiencing a housing bubble that is ripe to burst.

The development of housing prices is a hot topic for the media, analysts as well as the people in general. The reason for this is fairly obvious. It is said that around 95% of the population in Norway will some time in their life own their own dwelling (Eiendom Norge, 2013). Since the purchase of a dwelling is most likely the largest investment one will make, it is understandable that the housing market receives so much attention. It is almost viewed as a human right to own your own dwelling in Norway, and therefore the governing body in Norway will do what they can to keep it as attractive as possible to own your own dwelling. With 82% of the population over 16 years old owning a dwelling, Norway is amongst the very top in the world (SSB, 2016a). However, although the Norwegian government continues to upkeep the benefits, there are many regulations one must overcome before even being able to purchase a dwelling, making the market very stable compared to other countries. Considering all these aspects and the importance of a well-functioning housing market, this lays the foundation for an extremely interesting thesis for us.

1.1. Motivation Our master thesis is a long investigation about the current market conditions for housing in Norway. This is a topic that has received more and more attention in recent years as many people believe what we are experiencing now in housing prices cannot be fundamentally backed up.

Our motivation with this master thesis is to use the tools we have available and what we have learned through our master’s degree to give analytic and descriptive feedback regarding our topic. We will also use previously made models and try to use the data we have available today to see if they are still

5 applicable for today’s market conditions. Also, we wish to investigate the different driver’s effect on housing prices and how this has changed.

Both of us are students who are just about to start new jobs in Oslo. Therefore, the process and findings of this master thesis will to a large degree affect us in the near future as we are both planning to purchase an apartment. This clearly creates an even larger motivation for us to find out if we are planning to buy at a price peak, or if we can safely invest our money in housing for the years to come.

1.2. Problem Statement As stated in the beginning of this section, Norway has experienced a continuous growth in housing prices the last 25 years. It seems as if the longer this development continuous the more people are going to write, discuss and analyze whether we are experiencing a housing bubble or not. This can to some extent become a self-fulfilling prophecy as they will only continue to state this until one day housing prices will decrease for a longer period. Our goal is therefore to investigate if these statements about today’s housing market are accurate. This has led us to the following problem statement:

“Are we currently experiencing a bubble in the Norwegian housing market?”

To be able to reach the conclusion of this problem statement we will answer other sub-questions throughout the thesis:

➢ Is the housing price level and development unique for Norway when looking at other comparable countries? ➢ Is it possible for the fundamental factors on the supply and demand side of the housing market to explain the recent price growth? ➢ Can previously made housing price models precisely predict and explain the current market situation?

6 1.3. Methodology In this thesis, we have mostly used a quantitative research approach. Quantitative research is an approach for testing objective theories by examining the relationship among variables (Creswell, 2014). However, in the fundamental analysis we utilize a mixed approach. In a mixed approach, a mix between qualitative and quantitative research design is used. We will use quantitative research data, but analyze it in a more qualitative way where we make interpretations of the meaning of the data. The worldview and framework for conducting the thesis stems from a post-positivistic approach. By worldview, we mean the basic set of beliefs that guide action. A post-positivistic approach assumes an objective reality, but the absolute truth about it can never be found. Evidence from research is always imperfect and fallible. Thus, we never confirm a hypothesis, we can only reject (or fail to reject) the alternative hypothesis. Being objective is another key assumption of this approach. Validity and reliability in the data is therefore of high importance. We have a deductive approach to our analysis, where we have specific theories that we want to test with the collected data in order to test our hypothesis about the Norwegian market. This thesis does not seek to create new theories or models, but more to analyze and discuss the possibility of a housing bubble in Norway, and the drivers behind the increase. We therefore have a descriptive approach in this thesis, where we seek to describe more than explain.

1.4. Delimitation The Norwegian housing market can be considered very regional. This is to a large extent because of the extreme urbanization in Norway. Areas with large cities will in general experience a higher volatility in housing prices than rural areas. Also, if one region is highly correlated with an industry it will be more affected here than other areas. For our master thesis, we have chosen to look at the housing market as one. We will refer to certain regional examples, however the overall thesis will focus on Norway as one housing market. This is also true for dwelling types, when referring to “housing” and “house” it will entail houses, apartments, and other residences. Both these assumptions are indeed a simplified version of reality; however, we believe that it will not affect our final conclusion.

The time horizon we have chosen to focus on is primarily from 1980 until today. In the historical section, we will go further back, as this is only to get a picture of the historical development. However, all

7 analysis will be conducted using data from at the latest 1980. This is true for Norway and for the other comparative countries. For the comparable countries data availability has set the frame for how far back we could go.

The latest collected data we have chosen to include dates back to May 1st, 2017. All data and articles published beyond this date have not been considered.

1.5. Data The statistical and theoretical background for this thesis is based on secondary data. Our topic is a common subject for articles, theses and literature. The Financial Crisis of 2007 that erupted due to a housing bubble in the United States sparked even more discussion and published research on the topic. Not only does this make it easier to find information and data, but also to find good and reliable data. Using primary data would in this case neither be possible nor feasible, as we are not able to get better and more reliable data than what a large and well-known organization can. As we have used data from renowned statistics banks and researchers for our statistical tests, it makes us confident that our data has both high validity and reliability.

A weakness that comes from the plethora of articles and sources on our topic of interest is that it becomes harder to differentiate the reliable sources from the unreliable ones. We have therefore chosen to stick to well-known authorities on the subject for the theoretical foundation, as well as the dataset for the re- estimation. Most of the data used for our analyses is gathered from well-known statistical databases such as Statistics Norway (SSB), Statistics Sweden (SCB), Statistics Denmark (DST), Husbanken and the Norwegian Central Bank (NCB). Using these sources, we get stronger and more reliable results from our tests. We also back up our results and discussion with comments and thoughts from leading industry experts, including analysts, real estate agents and economists. This helps us see the current situation from multiple angles, as their different backgrounds makes them see the situation with different eyes.

For the re-estimation of Jacobsen and Naug’s house price model, we got the updated time series directly from Bjørn E. Naug himself (Naug, 2017). In the email, he also told us that some of the time series’ have

8 been revised since Jacobsen and Naug did their analysis back in 2004, so we will not be able to replicate their results. Getting data directly from NCB and Bjørn E. Naug strengthens our analysis, and it also helps us avoid mistakes that can happen when transforming or changing data to make it fit to a specific model or time period.

The Norwegian house price index has been gathered from NCB, and thus is the most reliable measure of the historic price development in Norway. In order to find the real house price development, we deflate it with the consumer price index (CPI), also downloaded from NCB.

In order to see the Norwegian market in comparison with other markets, we have chosen to compare some key figures with other markets. Due to the differences in the housing markets across borders the most relatable countries are the other Scandinavian countries, Sweden and Denmark. These countries are similar to Norway in population, culture, tax structure and the credit market. This way we can get a good indication to whether the Norwegian house price development is abnormal compared to relatable markets.

1.6. Thesis Structure The structure of this thesis is build up by four distinct sections. First, we will take the reader through the historical development of the Norwegian housing market. From this we will introduce the theories behind market bubbles, as well as the macroeconomic theory behind supply and demand in the housing market. Second, we will conduct an empirical comparative analysis using three well-known housing price models. Each model is explained in detail before the empirical test is conducted. Third, we will investigate many of the fundamental factors on both the supply and demand side of housing which are used in the various housing price models. From this we will look at how each of these fundamental factors can affect the housing market. In this section, we will also see how the fundamental factors in Jacobsen and Naug’s model have changed. In the end, we will tie all conclusions from the conducted analyzes together and interpret the results as well as state our final conclusion.

9 2. Historical Development of the Norwegian Housing Market To understand the different aspects of today’s market, we believe it is helpful to analyze the historical context. By looking at the Norwegian housing market from a historical point of view, we seek to identify both normal and abnormal periods. Using data from the NCB we can analyze the real prices from the time the first prices were recorded by the NCB in 1819, all the way up until 2015. As the NCB has not published the CPI for 2016, we cannot calculate the real house price for 2016. Comparing past situations to the one we have today is believed to make our conclusions stronger, as we have a historical perspective and backing behind our conclusions.

At the end of the 19th century, Oslo (at the time called Kristiania) experienced a population-boom as more and more people moved to the capital in order to gain from the higher wages in the city. All these new inhabitants needed housing, which resulted in an overestimated building-boom. After a few years this led to over 5000 empty dwellings in the capital. Deregulation in the lending market coupled with a massive demand for housing and mortgages resulted in no less than six new banks in 1897 alone. These banks specialized in mortgages with stocks as security and loans to construction firms, and they offered their clients liberal conditions. The rapid growth soon turned to speculation, which led to a bubble that burst on June 11th, 1899. All six banks went bankrupt in the years following the crash. The stock market crash had a direct impact on the housing market as a lot of the mortgages were directly tied to the stock market (Søbye, 2000). After 5 years, the market was again back to more normal price levels, but it would take almost 100 years before the real price of houses again returned to the level it was in 1899, which is found in figure 1.1. Real prices are nominal prices adjusted for inflation, which is done to make house prices comparable over time.

As the First World War came to an end, the Norwegian market experienced a boom in demand for consumer goods. Suppliers were not able to meet the new level of demand, which resulted in a massive price increase, high inflation, trade deficits, currency depreciation and an overheated economy (Grytten, 2002).

10 Figure 1.1

Real house price index 350

300

250

200

150

100

50

0 1819 1824 1829 1834 1839 1844 1849 1854 1859 1864 1869 1874 1879 1884 1889 1894 1899 1904 1909 1914 1919 1924 1929 1934 1939 1944 1949 1954 1959 1964 1969 1974 1979 1984 1989 1994 1999 2004 2009 2014

Source: NCB 2017; Own calculations

The Great Depression hit Norway harder than most countries. As a small open economy, it was affected more heavily by changes in trading partners, especially Great Britain and Sweden (Grytten, 2002). The opposite monetary policies introduced during the war and right before the Great Depression also made the situation worse. High inflation coupled with a banking crisis that led to a shortage in mortgages, caused a drop in the price growth for the housing market.

After World War II, the housing market stabilized and saw little growth in real prices. Structural changes to the credit market, more generous lending policies and a less regulated housing market all contributed to a tremendous rise in housing prices during the 1980's. In order to control the credit expansion, stricter regulations were introduced at the beginning of 1986. A drop in the oil price and an increased key rate led to a notable drop in housing prices, resulting in heavy losses for the banks. The stock market also went through a recession at that time, making the situation even worse. After the bubble burst and Norway went into a recession, housing prices decreased until it reached the bottom in 1992 (Torsvik, 1999).

11 The growth in the real estate market has been outstanding ever since prices started increasing again in 1992. The Financial Crisis only put a slight stop to the phenomenal growth in prices, and even though the government has tried to regulate the market and experts have said every year that the growth is going to stop, 2016 saw the largest growth yet with over 12.3% increase in total, and a 24% increase in Oslo. The only region that seems to be heavily affected by the low oil prices is Stavanger, the oil capital of Norway.

Figure 1.2

60000 Real price per square meter

50000

40000

30000

20000

10000

0 1819 1824 1829 1834 1839 1844 1849 1854 1859 1864 1869 1874 1879 1884 1889 1894 1899 1904 1909 1914 1919 1924 1929 1934 1939 1944 1949 1954 1959 1964 1969 1974 1979 1984 1989 1994 1999 2004 2009 2014 Oslo Bergen Trondheim Kristiansand Source: NCB 2017 (2015 prices); Own calculations

The largest cities in Norway the last 200 years have experienced different development in prices, this can be seen in figure 1.2 above. Oslo has throughout the period had the highest price per square meter, and the Kristiania Crisis is clearly visible with the peak in 1899. For Oslo, it took 106 years before prices per square meter again rose above the real prices in 1899. After 1950 it seems like the cities started to have more similar developments, and that they more started following the same trends. During the late 70’s the price difference per square meter between the different cities decreased, giving more similar price levels. This was mainly caused by an increase in prices in Bergen, Trondheim and Kristiansand. However, one can see that within the last ten years Oslo has skyrocketed compared to the other cities.

12 Comparing the average square meter price of Oslo to the average square meter price of Bergen, Trondheim and Kristiansand, the gap becomes evident. From 1992 to 2005 the average difference was about 6,044 NOK, while from 2006 until 2015 it became more than twice that, about 12,329 NOK. Kristiansand, being geographically close to Stavanger, felt the repercussions from the oil crisis that hit Stavanger hard, and therefore experienced a fall in real prices in 2014/2015.

The incredible growth in prices has made it lucrative for existing house owners, while young people looking to establish themselves in the market have experienced greater and greater difficulties. With housing prices, both in real and absolute terms being the highest they have ever been, the debate is raging whether these high prices can be explained by fundamental factors in the economy and whether the price level is sustainable. Will we continue to see record-breaking growth, or will we experience a bursting bubble? History has shown us that periods of high growth most often are followed by periods of declining prices.

13 3. Bubble Theory Defining a financial market situation as a "bubble" has no clear and unified definition. One definition that is used by this thesis and other analyses is the one put forward by Stiglitz (1990):

"If the price is high today only because investors believe the selling price will be high tomorrow, when the fundamental factors do not seem to justify such a price, then a bubble exists." (Stiglitz, p. 13, 1990)

Following the theoretical foundation of Grytten (2009), a housing bubble, or any speculative bubble, can be expressed like this:

Equation 3.1 1 � = ( ) � (� ) � 1 + � � �+1

�� is the value of the bubble, t is time, �� is expectations and r is the expected rate of return.

The equilibrium condition in a financial market can be written as:

Equation 3.2 1 � = ( ) � (� + � ) � 1 + � � �+1 (�+1)

This expression shows that price p for period t equals expected E rate of return d plus expected price on the financial object in the next period t+1, discounted with the required rate of return r. Over time, the price on the financial object will accumulate in line with this expression:

Equation 3.3 � 1 � 1 � � = ∑ ( ) � (� + 1) + ( ) � (� ) � 1 + � � � 1 + � � �+� �=1

14

The first part of the equation is the sum of discounted expected return for the entire period, while the other part of the equation shows the expected price at the end of the period. Present value of the price on the financial object will therefore be:

Equation 3.4 � 1 � � = ∑ ( ) � (� ) + � � 1 + � � �+1 � �=1

�� is a stochastic process satisfying equation 3.1. We get the following equation explaining the value of a bubble by rearranging equation 3.4:

Equation 3.5 ∞ 1 j b = � − ∑ ( ) E (d ) t � 1 + r t t+j �=1

Through equation 3.5 we see that the value of a bubble is the asset’s market price minus the discounted sum of future returns, or the assets fundamental value. If the fundamental value is less than the market price of the asset, there is a positive bubble value. As both the annual return on housing and the capital gain in previous periods are unknown sizes, the fundamental value is a relative theoretical term that has to be estimated. Comparing market value with the fundamental value is a common theme in bubble theory. This will therefore be tested in several ways throughout this thesis.

3.1. Case & Shiller’s Market Characteristics A bubble most often refers to a situation where public expectations of a future increase in price causes the price to be temporarily and artificially high. With housing bubbles, people tend to spend more and save less, as they believe they will be duly compensated by the expected price increase in the future.

15 They see their investment in real estate as a way of saving, as the price increase in itself is considered saving.

Case & Shiller (2004) use seven market characteristics to analyze whether a housing bubble is present. These characteristics are based on surveys they performed in 1988 and 2003. Seeing as these characteristics predicted both the bubble at the end of the 1980's and the Financial Crisis, we see them as useful indicators for predicting bubbles. Case and Shiller’s characteristics are: • People see housing as an investment • Widespread agreement of continuing rise in prices • Exaggerated Expectations, Excitement, and Word of Mouth • Sense of urgency in buying a home • Simple (or simplistic) theories • The occurrence of sales above asking prices • Perception of risk

3.2. Euphoric and Non-Euphoric Bubbles Grytten differentiates between two main types of bubbles: euphoric and non-euphoric bubbles. Euphoric bubbles are closely related to the seven characteristics, as these bubbles are driven by psychological factors in the market. The market participants believe that the prices will increase and therefore they are willing to spend more in order to enter the market. Non-euphoric bubbles are driven by fundamental factors, but the growth is not stable in the long run. The Kristiania Crisis is for example best described as a non-euphoric crisis. People back then used many of the same factors that are being used today in order to explain the price increase; urbanization, housing demand and increased income. Still the market crashed hard, and it took a long time for it to fully recover. Grytten also points at four factors that can help explain the recent increase in prices, but that are not stable in the long run: • Not enough building • Strong pressure from work immigration • Low interest rate • Low unemployment

16

"I believe this is a long-lasting state of emergency, and therefore people see this as the new normal. But I don't think this is normal." – Ola H. Grytten (Dagens Næringsliv, 2016a).

Speculation is often a clear indication of a bubble. When people buy to invest and not to live, it causes prices to increase even further. This also attracts investors with more money than those looking for a place to live, which in turn intensifies the competition for the more affordable dwellings. As the market in Oslo is the one that has experienced the most extreme increase, the Norwegian government has tried to stop these speculative investors by introducing new (temporary) regulations. Until June 2018, there is a 40% equity demand when buying a second dwelling in Oslo, and banks have a lot less freedom to give loans above 5 times total debt ratio (Regjeringen, 2016).

17 4. Supply and Demand in the Housing Market In this section, we will investigate the supply and demand theory of the housing market, which ultimately lays down the basis of how housing prices are determined in theory. It is important to distinguish between the short and the long-run when looking at the supply in the housing market, therefore we will present both situations and how the equilibrium is found. By using the theoretical procedure set by Jacobsen and Naug (2004) and Hendry (1984) in their articles, we will be able to shed light upon some of the explanatory factors within the supply and demand theory in housing.

4.1. Supply and Demand Theory The law of supply and demand is a fundamental economic principle that applies to all products. The principle states that, holding all else equal, if the supply of a product increases, the price will decrease. If the supply of a product decreases, the price will increase. Alternatively, if the demand of a product increases, the price will increase. If the demand of a product decreases, the price will decrease.

4.2. Supply and Demand in the Housing Market With regards to the housing market the supply is the housing stock available and the demand is the consumers who are looking to buy. The supply can be considered relatively flat, or fixed, in the short- run, as it takes time for new builds to be completed and the amount of construction per year is low considering the total housing supply. The supply could therefore be considered inelastic in the short-run. An oversupply in housing would result in what is called a “buyer’s market”, as the buyers will have many objects to choose from, which will ultimately drive prices down. An undersupply in housing would alternatively result in a “seller’s market”, where the supply of housing is scarce and the consumers fight to “win” each object available. This would result in a price increase.

The demand is continuously changing to the consumer’s preferences, which makes the demand elastic. A large increase or decrease in the consumer’s demand will therefore affect housing prices quickly and drastically. Consequently, in the short-run, housing prices fluctuate with the demand. In the medium to long-run, is it assumed that the supply will adapt to the demand, reaching its equilibrium.

18 4.2.1. Supply The supply within the housing market consists of two factors, (1) depopulation, demolition or renovation and (2) new builds. These two factors can be considered the total housing supply. New builds only comprise of about one percent of the total housing mass available each year and therefore have a very little effect of the supply curve in the short-run (NOU, 2002). New builds take time to complete mainly because of all the preparation that comes into play for the build to start. The more tangible aspects are for example land acquisition, workforce, materials, machines and tools, but also more importantly is the bureaucracy aspect which includes city council approval to build and other regulations. Because of all these factors the supply is generally deemed fixed in the short-run, or in other words perfectly inelastic. However, in the very long-run we can assume that real estate developers will adapt to the demand of the market and supply what is needed, therefore in this time horizon we can assume the supply is perfectly elastic. Worth mentioning is that vacant land in the big cities is scarce and in rural areas less so, but still not infinite, therefore this assumption of perfect elasticity is not theoretically sound.

Figure 4.1 Figure 4.2

Short-run equilibrium Long-run equilibrium Source: Own creation Source: Own creation

As written above, the housing supply can be viewed in the short-run and in the long-run. To describe the housing supply both in the short and long-run we can use Hendry’s model (Hendry, 1984).

19 Equation 4.1

�� = (1 − ��)��−1 + ��

Where,

�� = Housing supply, period t � = Depreciation rate of present housing supply

��−1 = Housing supply, period t-1

�� = Number of new builds, period t

As we can see from equation 4.1, the housing supply is described as the housing supply in the previous period ��−1, adjusted for the depreciation (1 − ��)�� and the number of new builds ��. As concluded by Hendry, since �� is small relative to the total housing supply, it is assumed that the housing supply is fixed in the short-run. In other words, the housing supply is constant to the previous period ��−1 (Rødseth, 1987).

4.2.1.1. Supply in the Short-run (Short-run equilibrium) With regards to what is written above, by definition, in a perfect market the equilibrium price for housing is found where the supply and demand curve intersect. Therefore, in the short-run when the supply curve is completely inelastic, the price is adjusted by changes in the demand. In figure 4.3, we can see how the supply is fixed in the short-run at point ��−1. The demand curve, defined as �1, intersects with ��−1 and creates an equilibrium price at �1. As the price is only subject to change from the demand in the short- run, we can see that a shock where demand goes to �2, makes the new equilibrium price go up to �2. Here it is easy to see that any change to the demand, either up or down, will have large effects in the short-run.

20 Figure 4.3

Source: Own creation

4.2.1.2. Supply in the Long-run (Long-run equilibrium) Following an increase in demand for housing, it is natural to assume that real estate developers will supply more housing for the market, as the market is willing to pay more for each unit. Therefore, within the medium- to long-run, new supply has been released to the market. Over the long-run horizon, one could assume the supply to be fully elastic and horizontal. According to Hendry, the long-run equilibrium could therefore correspond to a steady state in which real estate developers are earning normal profits and new builds exactly match the depreciation in the housing supply (Hendry, 1984). As the supply curve is horizontal, any price increase within the housing market will correspond to the inflation, holding all else equal.

∗ As seen in figure 4.4, the price, � , stays constant because the supply curve, ���, is horizontal in the long-run. This is shown as the demand, �1, moves to �2. The price stays the same.

21 Figure 4.4

Source: Own creation

4.2.2. Demand As written many times above, in the short-run the demand in the market will always be the main reason for price movements for housing. In the following section, we will use Jacobsen and Naug’s aggregate demand model to explain the demand side of the housing market.

First, it is important to distinguish between the two types of demand as explained by Jacobsen and Naug (Jacobsen & Naug, 2004). 1) Household demand for owner-occupied housing, i.e. demand for housing with the intention to live in. 2) Demand for housing as a pure investment instrument, i.e. demand where the sole goal is future capital gains either in the form of rent or by realized gain by sale.

We can assume that the first group is much larger than the second, and therefore it is the main focus both for Jacobsen and Naug’s article and our thesis. (Jacobsen & Naug, 2004).

The aggregate demand function is as follows,

22 Equation 4.2 � � �� = �( , , �, �) � ��

�1 < 0, �2 < 0, �3 > 0, where

�� = Housing demand V = Total housing costs for the general owner P = Index of prices for goods and services other than housing HL = Total housing costs for a general tenant (rent) Y = Households’ real disposable income X = Vector of other fundamentals that affect housing demand

�� = The derivative of �(*) with respect to agreement i

From equation 4.2 we can conclude that the demand for group number one will increase if income increases and will decrease if housing costs of ownership increase in relation to house rents or prices. According to Jacobsen & Naug, “the vector X represents various observable variables which capture the effects of demographic conditions, banks’ lending policies and household expectations concerning future income and housing costs” (Jacobsen & Naug, p. 31, 2004). This vector will be further explained later in this section.

The next equation explains the housing cost for a dwelling owner. “The housing cost measures the value of goods in which the owner relinquishes by owning and living in his own dwelling” (Jacobsen & Naug, p. 31, 2004).

The real housing price for owners (�/�) may be defined as:

23 Equation 4.3 � �� �� ≡ �� = [�(1 − �) − �� − (���� − ��)] � � � where BK = Housing cost per real krone (NOK) invested in a dwelling PH = Price for an average dwelling (in NOK) � = Nominal interest rate � = Marginal tax rate on capital income and expenses �� = Expected inflation (expected rise in P and HL, measured as a rate) ���� = Expected rise in PH (measured as a rate)

The first expression in the bracket [�(1 − �) − ��] shows the real after-tax interest rate. In other words, this is the direct cost of a mortgage. Jacobsen and Naug explains it as follows, “It measures the real interest costs associated with a housing loan and the real interest income lost by investing in a house” (Jacobsen & Naug, p. 31, 2004). From this we can see that an increase in the real interest rate will both increase the interest cost and the return when money is deposited in the bank, which ultimately increases the cost of living and decreases the demand. The second expression in the bracket [��� − ��] shows the expected real dwelling price growth. If this expression increases, the real housing costs will fall and conclusively the expected dwelling wealth will increase. From this one can say that relatively speaking, it will become more beneficial to own a dwelling than renting, and demand for housing increases. To sum equation 4.3 up, it shows the difference between the real interest rate after tax and the real price increase for housing. Jacobsen and Naug further simplify equation 4.3 into the following equation:

Equation 4.3* � �� �� = �� = [�(1 − �) − ����] � � �

Looking back at equation 4.3 once again, Jacobsen and Naug have also created a function for the variable Y, which represent the real disposable income. The equation for Y is as follows:

24 Equation 4.4 �� � = ��1���2���3

�1 + �2 + �3 = 1, �1 < �1, �2 < �2, where YN = Nominal disposable income

From observing equation 4.4, we can see that there are three factors in the denominator that will reduce the purchasing power of households and conclusively the demand. These factors are as mentioned;

P = Index of prices for goods and services other than housing HL = Total housing costs for a typical tenant (rent) PH = Price for an average dwelling (in NOK)

The last term of the aggregate demand function 4.2 of Jacobsen and Naug is the variable X. As stated above the vector X represents various observable variables which capture the effects of demographic conditions. Jacobsen and Naug show to examples such as migrations patterns, population size and strong urbanization, as significant demographic factors which can increase the demand for housing (Jacobsen & Naug, 2004). One of the most powerful explanatory factors in the variable X explains the impact of the banks’ lending policies. As most dwelling purchases are financed through mortgages the availability of credit for the consumer will without doubt affect the demand. Especially in Norway, where in 2016 the Financial Supervisory Authority of Norway (FSAN) presented numbers which stated that Norway had a ratio of household debt to disposable income of 297% (FSAN, 2016a). As concluded previously, lower interest rates will affect the housing demand in a positive way. Since the bank’s credit offerings can have such an impact on the housing market, Jacobsen and Naug created a function for this factor as well.

The banks credit offerings to households (�2) is as follows:

25 Equation 4.5 �� � = ℎ [�, ���, �, �, ] � �

ℎ1 > 0, ℎ2 < 0, ℎ3 > 0, ℎ4 < 0, ℎ5 > 0, where

�� = Bank’s supply of credit to households � = Bank’s profitability ��� = Measure of government regulation of bank lending � = Unemployment rate

ℎ� = The derivative of h(*) with respect to argument �

From observing equation 4.5 we can see that the bank’s credit offering will decline with stricter regulations, if the bank’s profitability decreases, and if there is an increase of unemployment. An increase in unemployment will ultimately decrease the expectation of future income and solvency.

26 5. Comparative Empirical Analysis In this section, we will use well-known models such as the Hodrick-Prescott filter, Price-to-Rent ratio, and Tobin’s Q to investigate the current development in housing prices in Norway. Using these results, we will be able to better conclude whether the Norwegian housing market is in fact facing a housing bubble or not. Also, to gain a better understanding of the Norwegian housing market we will look at Sweden and Denmark to compare our results. This is to further conclude if there are any abnormalities in the price development.

5.1. Hodrick-Prescott Filter The Hodrick-Prescott filter (HP filter) was first proposed by E.T. Whittaker in 1923, however it was further developed by Hodrick and Prescott in the 1980’s (Hodrick & Prescott, 1981). It quickly became a popular tool in the field of economics and is for example used by the Norwegian government and NCB (NCB, 2013). The HP filter is a mathematical tool used to remove the cyclical component of a time series from the raw data (Hodrick & Prescott, 1997). In a memo written by NCB they explain the use of the HP filter as follows, “the basic idea is that when the deviation between the indicator and trend is large i.e. the cycle is high, this may signal a financial crisis a few years ahead and should therefore trigger a response from policymakers to increase banks’ resilience to adverse shocks” (NCB, p. 2, 2013).

We assume that the original series �� is composed of a trend component (��) and a cyclical component

(��). This is presented in equation 5.1:

Equation 5.1

�� = �� + ��, � = 1, 2, 3 … , �

Hodrick and Prescott suggest a way to isolate �� from �� by the following minimization equation;

Equation 5.2 � �−1 2 2 ��� = {min ∑(�� − ��) + � ∑[(��+1 − ��) − (�� − ��−1)] } , � = 1, 2, 3 … , � �=1 �=2

27 The residual �� − �� (the deviation from the trend) is commonly referred to as the business cycle component. The deviations are squared to give equal weights to both positive and negative deviations. The second part of the equation measures the change in the trend from one period to the next and includes the smoothing parameter �, which penalizes the acceleration in the trend relative to the business cycle component (Ravn & Uhlig, 2002). As � approaches 0, the trend component becomes equivalent to the original series. This would be considered the optimal condition, as the deviation between the actual data and trend is zero. However, this is highly unrealistic as it would imply that business cycles do not exist. When � gets close to infinity, the trend component approaches the linear trend. (Kim, 2004). This is also unrealistic, as this implies that the trend is 100% linear with constant growth.

One of the most discussed aspects of the HP filter is the determination of the smoothing parameter �. Many choose to follow Hodrick and Prescott’s value of 1600, however this value is mostly used for quarterly data (Hodrick & Prescott, 1997). The question that arises is, what value should one use when looking at annual data. Backus and Kehoe (1992) use a value of 100, where Correia, Neves, and Rebelo (1992) suggest a value of 400. Even more recent is the conclusion from (Ravn and Uhlig, 2002), where 1600 they suggest a � of 6.25, which is found by . 44

5.1.1. Limitations Although the HP filter is widely used in the world of economics, there are many potential weaknesses due the simplicity of the model. Here are some of the weaknesses that can be mentioned:

Choosing the smoothing parameter. As we just wrote above, there is no clear value that one must use for the parameter �. It is set subjectively and can affect the results of the model greatly. Studies have been made to find the best fitted smoothing parameter, however one can never be certain that the model will produce the best fitted trend for the time series.

Equal weight of up- and downturns. As there are both up- and downturns in an economy the HP filter must deal with these fluctuations in the best manner. However, the HP filter equally weights the up- and downturns. In other words, the model assumes that these economic states are equally long lasting.

28 However, research has concluded that this is not the case. Cristina Romer concluded in her research that economic upturns are longer lasting than economic downturns (Romer, 1999).

End-point problems. The HP filter uses previous, current, and future data points to determine the trend in a given time period. This is a problem for both end points, beginning and ending, as there are no data from before the start point and no data from after the end point. Hence, the trend at both end points will be estimated from current and future data and previous and current data, respectively. Consequently, the trend-estimates at these two end-points will be more affected by current observations than the rest of the series.

Real-time problems. This problem makes the last-mentioned end-point problem even more severe, and it is arguably the most critiqued aspect of the HP filter. It is well known that current data in time series are often uncertain, and can be changed after some time. Therefore, the end-point problem becomes even bigger as these trend-estimates are given more weight than the rest of the time series.

Professor at UC San Diego, James Hamilton has recently written an article where he heavily critiques the HP filter. His arguments are very much in line with the problems we have mentioned. His arguments are as follows: “(1) The HP filter produces series with spurious dynamic relations that have no basis in the underlying data-generating process. (2) A one-sided version of the filter reduces but does not eliminate spurious predictability and moreover produces series that do not have the properties sought by most potential users of the HP filter. (3) A statistical formation of the problem typically produces values for the smoothing parameter vastly at odds with common practice, e.g. a value for � far below 1600 for quarterly data” (Hamilton, p. 1, 2017).

5.1.2. Empirical Testing Before we can go forth with applying the HP filter we must conclude which smoothing parameter we will use for �. As stated above, Hodrick and Prescott concluded in their paper that a smoothing parameter of 1600 would give the best fit for their quarterly data. Our analysis considers annual data; therefore, we must take this into consideration. The annual data we have collected for our analysis goes back to 1819.

29 One of the reasons for why we have chosen to look so far back, rather than only looking from 1980 and out is to pick up the historical development. Also, another positive side of having so many data points is that we are able to eliminate some of the end-point problems that occur with the HP filter, at least from the start point. It will be difficult to properly analyze the current situation; however, we will be able to analyze previous bubbles and see whether the model fits or not. If previous bubbles are captured, then we can write about the results with higher confidence.

There are many aspects to think about before choosing the value for �. A level of 100 has been considered for annual data (Hodrick & Prescott, 1997). The recent development in the Norwegian housing market may pose a challenge for having a low � since the trend will follow the current extreme values, which in turn will underestimate a potential bubble. Another problem is highlighted by the European Central Bank, “the smoothing parameter does not only effect the cycle but the volatility of trend growth and well – a consequence of the fact that the HP filter does not contain an explicit model of the cycle.” (ECB, p. 9, 2005). This is why many economists argue to use high values for � when analyzing annual data, because they feel that when using lower values, it would give rise to implausible volatile trend growth rates. Also, using a higher smoothing parameter will ultimately provide more volatility, which also makes a larger portion of the fluctuations a result of temporary disturbances.

To be able to capture both aspects of previous academic conclusions we will use one low and one high smoothing constant. We have chosen to use a newer and an older conclusion of the best fitted value for �. We will use Ravn and Uhlig’s smoothing parameter for annual data of 6.25 and Correia, Neves, and Rebelo’s smoothing parameter for annual data of 400. Hence, we will be able to see some changes in the results and investigate both sides of the discussion of which value for � to use.

Figure 5.1 shows the development of the real house prices for Norway in general and both trend components, using 6.25 and 400 as the smoothing parameter from 1819-2015. When looking at the smoothing parameter of 6.25, the trend moves very close to the real house price, as expected. Therefore, the real house prices only show to be overpriced for a short window in five distinct time periods, which we have already previously have mentioned. That is, the Kristiania Crisis, World War I, the Great Depression, the Norwegian Baking Crisis, and the Financial Crisis. What seems to follow all these crises

30 is that the real house prices showed signs of being underpriced at the bottom of the crash. Interestingly enough, if we look at the recent development it actually seems as if real house prices are undervalued. In 2014, real house prices experienced a minor drop, which is not too surprising considering the recent oil price drop. Here it must be mentioned, that our data only includes data up to 2015, as NCB has not released the most recent data. Therefore, the model does not include 2016’s leap in housing prices.

Figure 5.1

Real house prices with trend lines, Norway (1819-2015) 350 300 250 200 150 100 50 0 1819 1834 1849 1864 1879 1894 1909 1924 1939 1954 1969 1984 1999 2014

Real House Prices HP - 6.25 HP - 400

Source: NCB, 2017; Own calculations

Next, we look at the smoothing parameter of 400. As the trend becomes more linear with constant growth as � increases, it becomes more evident to observe the financial turmoil’s that we have previously mentioned. It becomes clearer that real house prices have been overpriced in these periods, as well as underpriced when the bubbles hit bottom. The gap between the trend and the real house prices during the Norwegian Banking Crisis is the largest gap viewed in our data. After this, the trend seems to have followed the real house prices. Conclusively, given the data we have available, the model indicates that Norway is currently experiencing an underpricing.

The conclusion changes when looking at figure 5.2, which shows the development of the real house prices for Oslo, with both trend components using 6.25 and 400 for � from 1841-2015. The data for Oslo looks fairly similar to the data for Norway in general. It could be argued that the data is more volatile in

31 Oslo. Also, the magnitude of the Kristiania Crisis in 1899 is much larger, which is expected as this was primarily experienced in Oslo. The main characteristic we want to point out from this graph is the fact that both trends using smoothing parameters of 6.25 and 400 are below the real house prices from 2014 and out. From this we can conclude that both trends imply that the real house prices in Oslo are overpriced.

Figure 5.2

Real house prices with trend lines, Oslo (1841-2015) 350 300 250 200 150 100 50 0

Real House Prices HP - 6.25 HP - 400

Source: NCB, 2017; Own calculations

We have observed that we currently have two different scenarios in Norway. For Norway in general today’s real house prices are below the estimated HP filters, both when using smoothing parameter’s � = 6.25 and � = 400, suggesting undervalued prices. However, when looking at Oslo by itself, the real house prices are above the estimated HP filter, which implies overvalued housing prices. From this we can conclude that there could exist a bubble in Oslo, but not for Norway in general.

5.1.3. Comparison To be able to fully evaluate the results for Norway it would be beneficial to have some relevant comparisons. From this one could with more confidence conclude whether the results from Norway are abnormal. As previously mentioned we will look to the other Scandinavian countries.

32 The first country we will use the HP filter and compare the results from Norway with is Sweden. As the historical data available for Sweden is much shorter than Norway the time horizon here will be 1975- 2016. Nonetheless, it is within this period the largest growth rate has occurred for all Scandinavian countries. Therefore, this window should be sufficient enough to be able to conduct the analysis with good results.

Sweden Sweden went through its own banking crisis as Norway, between 1990-1995 (NCB, 2011). This is clearly seen in figure 5.3 which shows that the housing market was overpriced and not long after corrected to its average mean. Since then the housing market has grown, but the growth rate has not surpassed the trend line significantly. There are two periods where this happened and it was during the Financial Crisis, and now in recent times. Therefore, unlike Norway, one can conclude that given this data the Swedish housing market is in general overpriced.

Figure 5.3

Real house prices with trend lines, Sweden (1975-2016) 300

250

200

150

100

50

0 1975 1980 1985 1990 1995 2000 2005 2010 2015

Real House Prices HP - 6.25 HP - 400

Source: SCB, 2017b; Own calculations

Next, we look to the capital of Sweden, Stockholm. We did this for Norway as well to see if there were any deviations between the situation in the entire country and in the capital. In Norway, this happened to be the case. Looking at figure 5.4 the same trends are seen in Stockholm as in Sweden in general.

33 Interestingly enough, it looks like the Financial Crisis did not have as great of an effect in Stockholm as it did for Sweden in general. If we look at the current market situation, the housing market in Stockholm seems to be even more overpriced and further away from the trend lines than in Sweden in general. This is also in line with recently published articles which state that Stockholm currently is experiencing record high prices per square meter of 94.000 SEK (SvD Naringsliv, 2017).

Figure 5.4

Real house prices with trend lines, Stockholm (1975-2016)

400 350 300 250 200 150 100 50 0 1975 1980 1985 1990 1995 2000 2005 2010 2015

Real House Prices HP - 6.25 HP - 400

Source: SCB, 2017b; Own calculations

Denmark Unfortunately, the only data we have been able to retrieve for Danish housing prices goes back to 1992. Therefore, the Danish time series is both shorter than the Norwegian and the Swedish time series. However, as we argued for the Swedish times series our main interest is recent developments, hence, there is no need to go as far back as the Norwegian time series. What is interesting to investigate is how the Danish housing market is doing compared to the Financial Crisis period, where the Danish housing market was hit particularly hard, compared to the other Scandinavian countries. The Danish housing market slumped around 30% through 2009 and only recently got back to the same housing prices level (Bloomberg, 2016).

34 Many chief economists and analysts in Danish banks have raised concerns about the current Danish housing market, mostly because of the fact that Denmark is currently the country which has experienced negative key rates the longest. Chief analyst Tore Stamer at Nykredit in Copenhagen said, “To be concrete, there is a danger that Danes will go blind to the risk of rates ever rising again” (Bloomberg, 2016). From this it will be interesting to see how the current situation is for the housing market in Denmark in general and Copenhagen.

When looking at figure 5.5 the Financial Crisis is very easy to spot. The housing market clearly experienced an overpricing compared to the long-term trend, and was eventually corrected. The housing market slumped so much that between 2010-2015 the housing market could have been considered underpriced. It is only in the past year the real house prices have yet again, only barely, moved above the long-term average for Denmark in general. This implies that Denmark in general is barely overpriced, using both smoothing constants.

Figure 5.5

Real house prices with trend lines, Denmark (1992-2016) 300 250 200 150 100 50 0

Real House Prices HP - 6.25 HP - 400

Source: Boligøkonomisk Videncenter, 2017; Own calculations

Now, if we look at figure 5.6 we can see the development for Copenhagen alone. The development of the figure is similar to Denmark in general. What is interesting to notice is the current deviation between the real house prices and both trend lines are larger than for Denmark in general. Therefore, the HP filter

35 implies that there is an overpricing in the Copenhagen housing market, arguably showing bubble tendencies. This is also in line with what the director for the national bank of Denmark, Lars Rohde, said in mid-2016, “We have seen a moderation of price developments this year compared to last year. However, it is still very strong for apartments. This is particularly true for Copenhagen’s housing market, which should be closely followed” (DR Nyheter, 2016).

Figure 5.6

Real house prices with trend lines, Copenhagen (1992-2016) 450 400 350 300 250 200 150 100 50 0

Real House Prices HP - 6.25 HP - 400

Source: Boligøkonomisk Videncenter, 2017; Own calculations

By looking at Sweden and Denmark in general as well as their respective capitals we are able to investigate and compare whether Norway is experiencing an abnormal housing market or not. As the mentioned Scandinavian countries are relatively similar there are good premises to compare them to each other.

In retrospect, we concluded that the HP filter implied that the Norwegian housing market in general was not overpriced, and hence does not show bubble tendencies. However, both Denmark and Sweden in general showed signs of overpricing, although Denmark showed very little. To sum this, Norway in general is, relative to the HP filter, closer to its long-term average compared to its counterparts. With that said, all capitals; Oslo, Stockholm and Copenhagen are showing signs of overpricing and hence,

36 bubble tendencies. The fundamental factors for why this could be happening in Norway will be covered in a later section.

5.2. Price-to-Rent Ratio The alternative to owning a dwelling is strictly speaking renting from another dwelling owner. The development between housing prices and rent prices is a relationship that has gotten more and more attention from economists in recent years. By looking at this relationship, one can try to identify a potential mispricing of dwellings relative to a long-run horizon and investigate whether there are bubble tendencies in the housing market. The Price-to-Rent ratio (P/R ratio) stems from the more known Price- to-Earnings ratio (P/E ratio) which is used to price stocks. The well-known P/E ratio was first developed by Gordon and Shapiro in 1956 (Gordon & Shapiro, 1956) and was further developed by Miller and Modigliani in 1961 (Miller & Modigliani, 1961). Both the real and the fundamental P/E ratio are presented in the equations below;

Equation 5.3 � ������ ����� ��� �ℎ��� ���� = � �������� ��� �ℎ��� (���)

Equation 5.4 � � ����������� = � � � (1 − �) ∑∞ � �=0 (1 + �)

The real P/E ratio calculates the value of a stock by dividing the current stock price (market price) by the earnings per share, as shown in equation 5.3 above.

The fundamental P/E ratio, which is found in equation 5.4, equals the sum of all discounted future dividends and the expected earnings in period t (��), deducted with the company’s retained earnings (�), discounted with the rate of return (�). The fundamental P/E ratio shows how much a stakeholder must pay for each unit of dividend he or she will receive in the future. One can identify a mispricing in the

37 market by comparing the real and the fundamental P/E ratio. If there is a deviation between the two, where the real P/E ratio is over the fundamental P/E ratio, one can argue for bubble tendencies.

The P/R ratio, which was developed by Poterba in 1984, is similar to the P/E ratio in the sense that it considers a dwelling price to be equal to the discounted value of all future profits tied to owning a dwelling (Poterba, 1984). By profit it is meant to be the value of the cost of owning and/or rental income, or alternatively the rent costs for similar housing. Poterba referred to the cost of owning a dwelling as;

Equation 5.5 ���� �� ������ � �������� = �(�� + � + � − �)

Where: � = Housing price index �� = Nominal interest rate after tax � = Property tax � = Other costs of owning (maintenance, risk premium, etc.) � = Expected capital gain

Equation 5.5 can be explained such as, the cost of owning is the sum of interest expenses given by owning a dwelling (this includes interest expenses that follows when owning a dwelling, but also interest income one gives up by locking up equity in housing), other costs of owning, minus the expected capital gain of owning a dwelling.

A rational dwelling owner will undertake a cost-benefit analysis of the dwelling he or she owns, where the benefit is the rent income one forgoes by living in the dwelling alone, and the cost is the cost of owning the dwelling (Poterba, 1984). In other words, in a long-run horizon the costs of owning a dwelling will be equal to the rent costs for a similar housing. This relationship is show in equation 5.6. below.

38 Equation 5.6 � = � (�� + � + � − �)

Where: � = Rent costs for a similar housing

From equation 5.6, we can see that it will be more beneficial to rent than to own if the rent costs are lower than the costs of owning, and vice versa. However, if these two are equal, one would be indifferent between renting and owning a dwelling. If the equation is not at its equilibrium, the demand will naturally go more towards the more economically beneficial choice for the consumers. This is in line with the general supply and demand theory we mentioned in section 4. From this we know that the price will be pushed up for the most economically beneficial alternative or down for the least economically beneficial alternative. A combination between the two will see to that the equation will reach its long-run equilibrium.

By rearranging equation 5.6. we are left with the fundamental P/R ratio;

Equation 5.7 � 1 ����������� = � (�� + � + � − �)

By rearranging the equation above, the right side of the equation shows that there is a long-run fundamental relationship between dwelling prices and rent prices, which in turn is the fundamental P/R ratio. We can find the fundamental P/R ratio by inserting the variables set from equation 5.5. From this we know that the fundamental relationship is a result of the interest rate after tax, property tax, other costs of owning a house, and the expected capital gain. All these variables will change over time, and therefore constantly change the fundamental relationship between dwelling prices and rent prices. For example, a lower interest rate will make it cheaper to finance a mortgage, which ultimately will make it more appealing for a household to buy a dwelling over renting and thus swing the demand more towards owning. Therefore, holding all else equal one can assume that the fundamental P/R ratio will go up,

39 when interest rates go down (NCB, 2003). On the other side, a lower expected capital gain on housing will make it less attractive to own, and thus swing the demand towards renting. From this we can be certain that the fundamental P/R ratio will not be constant in the long-run, hence an increasing fundamental P/R ratio does not necessarily mean that housing prices are increasing at an unsettling rate, or that there are bubble tendencies in the market. To able to say anything about this, it would be beneficial to compare the fundamental P/R ratio to the real P/R ratio in the housing market.

The real P/R ratio is found by dividing dwelling price over rent price such that each dwelling has its own P/R ratio. However, this is unpractical to use for a longer time horizon. Therefore, the most common way to find the real P/R ratio is to use the housing price index divided by a rent price index. The real P/R ratio can be found in equation 5.8 below.

Equation 5.8 � ������� ������ ����� ���� = � ������ ���� ������

The real P/R ratio has in recent years gained a lot of attention from the media and analysts, since it may provide useful information about the development in the housing market. Today’s P/R ratio is analyzed and compared to the historical average to see if the ratio is following a natural average path. If the path is not following a natural curve, one can start to question whether or not the ratio is at its equilibrium. What is important to remember is that media and other participants in the market often forget that the development of the P/R ratio can also be explained by changes in fundamental factors, and from this come up with conclusions on wrong premises. If housing prices have increased disproportional in relation to the rent prices, making the real P/R ratio unnaturally high, one could argue that the housing market is too hot and that there are bubble tendencies in the market. From this one could speculate whether the dwelling prices are driven by irrational expectations of the consumers with regards to future price movements. This is why we will apply the ratios to the market place to see if the bubble tendencies can be supported by the current data. First, we have to look at the assumptions of the model.

40 5.2.1. Assumptions As with most other models and theories there are certain limitations and assumptions that follow the P/R ratio. It is important to keep all of these assumptions in mind when we present the conclusions of our analysis.

It is assumed that all housing is homogenous, i.e. that there is a corresponding rent for each house. Underlining this is a strong assumption that the geographical location of housing does not affect the dwelling price or rent. Obviously, this assumption would never hold in the real world as we know for a fact that for all of Norway and more particularly Oslo, location has a great effect on the value of housing. Therefore, to apply the P/R ratio it is necessary to oversimplify and use aggregated numbers for many types of housing and rent.

Owning and renting are considered perfect substitutes. This assumption assumes that any increase in price for either side will immediately lead to a higher demand for the opposite side. This is a truth, with some modifications, as there is a certain lag in the housing market. Rent contracts are usually set for longer periods and take time to terminate, because of regulations for termination notices. Another aspect is the fact that many consumers have strong preferences for owning over renting, and can therefore not be considered to be in the same market. Just by looking at the own-to-rent ratio in Norway one can argue that both markets are not the same. In 2015, close to 80% of houses were occupied by owners, and around 20% were occupied by renters (SSB, 2016b).

Zero transaction costs. The theoretical framework for the model assumes no transaction costs for buying and selling housing. An example of violation of this assumption is found in the document duty in Norway. This regulation states that 2.5% of the purchasing price of full-owner housing must be paid to the Norwegian government (Kartverket, 2017). Clearly, this can be a significant cost for the buyer. Also, the costs related to searching for housing is assumed to be insignificant.

41 5.2.2. Data Material In the following sub-section, we will provide information about what data we have used for the calculation of the P/R ratio and where we have collected the data from.

Housing price index (P): For the house price data, we have used the annual housing price index, published from the NCB each year (NCB, 2017). From this data set we also have the price per square meter, going back to 1980. Annual Rent (R): As there is no statistical database which goes back to the start of our focus period of 1980-2016, we are forced to make some own calculations for this number. However, SSB started to provide statistical data for average rent per square meter from 2006 (SSB, 2017e). From this we can use the CPI data SSB has for the sub group of “paid rent”, which is an overview of an underlying sub group’s total contribution to the CPI (SSB, 2017d). Combining these two we are able to estimate the rent price per square meter from 2014 and back to 1980. We will also use this relationship to generate rent prices up to our end year, 2015. We will do this by taking basis in 2014 for the average price per square meter for rent, and taking basis in 2014 for the CPI as well, such that 2014 = 100.

Fundamental P/R ratio The Organization for Economic Co-operation and Development (OECD) conducted a similar study and used the fundamental P/R ratio for 18 OECD countries in 2006 (OECD, 2006). Therefore, we will use the same variables as they used for this study. The variables used are;

Mortgage rate: Data for average lending rates from banks is collected from SSB (SSB, 2017h) After tax mortgage rate: With the exception of the last two years in our study the tax rate has been 28% in Norway, therefore this is the number we have used (SSB, 2017a). The analysis conducted by the OECD use this value as well. Therefore; (i) * (1-0.28) = After tax mortgage rate. Costs of owning: We will use the same constant for holding cost as the OECD used for all countries, 4% (OECD, 2006). Property tax rate: Not all municipalities in Norway apply a property tax rate, however as of 2016 more than 85% did. Therefore, we will apply the highest possible tax rate for all, which is 0.7% (SSB, 2016c).

42 Capital Gain: The CPI is chosen as the parameter for capital gain. We used a five-year moving average for each year.

5.2.3. Empirical Testing In this sub-section, we will use the data discussed above to analyze the Norwegian housing market. We will do so by looking at the real and fundamental P/R ratios. First, we will look at the development of some of the underlying factors of the P/R ratio. This is to gain a better understanding of the development, and perhaps we will be able to see where certain factors have changed trajectory compared to the others. Second, we will look at the real P/R ratio and see if there is anything to learn from its development. Third, we will put both the real P/R ratio and the fundamental ratio in the same graph to see their change in relation between each other. From this we can conclude if there are any bubble tendencies in the Norwegian housing market.

In figure 5.7 below we can see the development of the CPI, rent prices, and house prices from 1980 until 2015. All factors have the same start base of 1, where 1980 = 1. By looking at the figure we can see that all three factors followed each other relatively close up until 1993 where they all converged. However, interestingly enough from this point and until today the rent prices and CPI have followed each other closely, whereas housing prices have heavily increased to over five times the value of 1993.

43 Figure 5.7

Underlying factors in the P/R ratio - Norway 14,0 12,0 10,0 8,0 6,0 4,0 2,0 0,0 1980 1985 1990 1995 2000 2005 2010 2015

CPI (1980 = 1) Rent (1980 = 1) House Price (1980 = 1)

Source: NCB, 2017; SSB, 2017e; Own calculations

As one can see from figure 5.8 below the real P/R has fluctuated throughout the last decades. The first large drop in the real P/R ratio seen during the Norwegian Banking Crisis at the end of the 1980’s. This is in line with the house price development, which we analyzed in the historical development in section 2. After the real P/R ratio hit a bottom low of 6.05 in 1992, the ratio has continued to increase until today. The Financial Crisis is definitely found in the figure, however, as we have concluded before, the crisis did not have that large of an effect on the Norwegian housing market one may have expected. Therefore, it did not take many years before the real P/R ratio was back at its pre-Financial Crisis level of 18.40 in 2011. Since then the ratio has continued to increase and was recorded at a record high of 19.81 in 2015.

44 Figure 5.8

Real P/R ratio - Norway 25,0

20,0

15,0

10,0

5,0

0,0 1980 1985 1990 1995 2000 2005 2010 2015

Real P/R ratio

Source: NCB, 2017; SSB, 2017e; Own calculations

Another aspect we can learn from this figure is the fact that the real P/R ratio is well above its long-term trend, or average of 11.76. When there is a large difference between the average trend and the actual real P/R ratio, it could be argued that there are bubble tendencies in the market. The assumption is further validated when one realizes that today’s real P/R ratio is almost twice the amount of the long-term average. From this we can conclude that housing prices are most likely overpriced. However, to say this with more confidence we must look at the relationship between the real P/R ratio and the fundamental ratio.

5.2.4. Fundamental vs. Real P/R Ratio Since the real P/R ratio has had its biggest development after 1990, we will only focus from 1990 and out when comparing the real and the fundamental P/R ratio. Both ratios have been converted into indexes, where 1990 = 1. The fundamental vs. real P/R ratio are found in figure 5.9 below.

45 Figure 5.9

Fundamental vs. Real P/R ratio - Norway 3,0

2,5

2,0

1,5

1,0

0,5

0,0 1990 1995 2000 2005 2010 2015

Fundamental P/R ratio (1980 = 1) Real P/R, (1980 = 1)

Source: NCB, 2017; SSB, 2017e; OECD, 2006; Own calculations

From figure 5.9 it can be seen that the fundamental ratio has shown tendencies of being more volatile than the real P/R ratio. As there are more factors that directly effect this ratio it makes more sense that it is in fact more volatile. By looking at the figure we can see that the real P/R ratio surpassed the fundamental ratio around 1998 and has been above ever since. What is interesting to notice is the fact how the gap between the two values has increased since 2010. Obviously, this only increases the validity of the conclusion that there may exist a bubble in the Norwegian housing market.

It must be mentioned that although the deviation between the two is large at the moment, there are certain factors that may back up why the fundamental P/R ratio is so low compared to the real P/R ratio. As written above, when there are certain shifts in the fundamental values of the ratio, it does not necessarily mean that there is a bubble in the market. For example, we have previously concluded that the key rate is at a historic low at the moment, therefore this will greatly affect the purchasing power of the participants in the market. Not only that, but when interest rates are low, landlords are able to charge less as they have lower costs tied to housing (Larsen, 2005). This will without doubt affect these two ratios. Nonetheless, the ratios are meant to give guidance with regards to the market condition, and as it stands now one could argue that there are bubble tendencies in the Norwegian housing market.

46 5.2.5. Limitations There are certain limitations one must keep in mind when analyzing these results. The collection of capital gains values and rent estimates are not perfect. If the capital gains parameter was more precise, one could have gotten a slightly different fundamental P/R ratio. Also, the rent index was estimated due to the fact there is no data that goes further back than 2006, which is too short of a time horizon for us to analyze and retrieve meaningful results. Therefore, there were some assumptions made and the rent prices where estimated back to 1980. Both these values were not perfect; however, we believe they gave a good picture of the actual situation in the Norwegian housing market.

Another aspect to think about is that although the fundamental P/R ratio takes into consideration many important factors regarding housing prices, there are still factors unaccounted for. Some of these other fundamental factors will be mentioned later in the thesis.

Lastly, one must not forget the assumptions we mentioned prior to conducting the analysis as well. All these assumptions were in some way violated, and therefore one cannot take the results as the actual situation. However, again, the analysis provides us with an understanding of the housing market, and points us in a direction. Therefore, we cannot use these results alone, but along with multiple analyses to see if there is a trend in the results.

5.2.6. Comparison To be able to gain a strong hold in the conclusion of the Norwegian housing market and its P/R ratios, it would be beneficial to compare its results to an equivalent country. By looking at recent housing price developments, as shown in figure 5.10, it is clear that Sweden and Norway historically have followed each other fairly close. Hence, this is another reason for why comparing these two markets could be an interesting study. Also, according to a recent study conducted by Barclays and the European Systematic Risk Board (ESRB) they ranked the risk for a collapse in the Swedish housing market as the highest for all European countries (ABCNyheter, 2017). From this we can investigate if our results from the P/R ratios match the experts’ prediction.

47 Figure 5.10

Source: HolbergFondene, 2017

Before applying the P/R ratios it must be said that the same assumptions previously mentioned apply for Sweden as well. Furthermore, as with Norway, all assumptions are violated for the same reasons.

5.2.7. Data Material – Sweden For Sweden, we were only able to retrieve sufficient data material back to 1996. Although this is not as far back as for Norway, we view the length of the time series sufficient enough to conduct the analysis. Next, we will go through where we have collected the needed material from.

Housing price index (P): We collected the housing price index data from Statistiska centralbyrån using the “Fastighetsprisindex” (SCB, 2017b). The price per square meter was collected from Svensk Mäklarstatistik using an average of Bostadsrätt and Villa (Svensk Maklarstatistik, 2017). Annual Rent (R): As there is no statistical database which goes to the start of our time period of 1996- 2016, we are forced to make some own calculations for this number. However, SCB provided statistical data for average rent per square meter in 2014 (SCB, 2015). From this we can use the CPI data SCB has for the sub group of “paid rent”, which is an overview of an underlying sub group’s total contribution to the CPI (SCB, 2017a). Combining these two we are able to estimate the rent price per square meter from

48 2014 and back to 1996. We will also use this relationship to generate rent prices up to our end year, 2016. We will do this by taking basis in 2014 for the average price per square meter for rent, and taking basis in 2014 for the CPI as well, such that 2014 = 100.

Fundamental P/R ratio The Organization for Economic Co-operation and Development (OECD) conducted a similar study and used the fundamental P/R ratio for 18 OECD countries in 2006 (OECD, 2006). Therefore, we will use the same variables as they used for this study. The variables used are;

Mortgage rate: Data for average lending rates from banks is collected from SCB (SCB, 2017c). After tax mortgage rate: The personal income tax rate for Sweden in the given time horizon has stayed around 30%, plus-minus a percent or two (Skatteverket, 2016). Therefore, we will use 30% as a proxy for the time horizon. Hence; (i) * (1-0.30) = After tax mortgage rate. Costs of owning: We will use the same constant for holding cost as the OECD used for all countries, 4% (OECD, 2006). Property tax rate: The property tax rate in Sweden in the given time period has stayed around 0.50%, and will therefore be used by us for the whole period (Skatteverket, 2016). Capital Gain: The CPI is chosen as the parameter for capital gain. We used a five-year moving average for each year.

5.2.8. Empirical Testing – Sweden First, we investigated the development between the CPI, rent prices, and house prices for Sweden to see how this has moved relative to each other and to Norway. The results proved to be very similar to Norway, in the sense that the CPI and rent prices has moved slightly upwards and close to each other, whereas the housing price index has moved further away from the other two factors. As with Norway, this development is not sustainable in the long run. Either the rate at which the house price index is increasing must come down, or the CPI and rent levels must come up. The result is presented in figure 5.11. below.

49 Figure 5.11

Underlying factors in the P/R ratio - Sweden 4,5 4,0 3,5 3,0 2,5 2,0 1,5 1,0 0,5 0,0 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

CPI (1996 = 1) Rent (1996 = 1) House Price (1996 = 1)

Source: SCB, 2015; Svensk Mäklarstatistik, 2017; Own calculations

The average P/R ratio for Sweden in its time horizon is 18.32. If we look at the development of the P/R ratio in figure 5.12, we can see that the current rate of 35 is almost double the rate of its average. In an article written by Time magazine in 2010 they implied that any ratio above 20 indicates that home prices are expensive relative to renting in the area (TIME, 2010). From that article one can assume that people in Sweden are paying far too much for housing compared to what they can get for renting. Further supporting the assumption that there are bubble tendencies in Sweden.

50 Figure 5.12

Real P/R ratio - Sweden 40,0 35,0 30,0 25,0 20,0 15,0 10,0 5,0 0,0 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

Real P/R ratio

Source: SCB, 2015; Svensk Mäklarstatistik, 2017; Own calculations

Lastly, we want to look at the development between the real P/R ratio and the fundamental P/R ratio. This will give us a better picture of how the housing market in Sweden is actually doing. In figure 5.13 we can see that the gap has increased since the beginning of the time horizon in 1996. Also, the rate of which it is increasing seems to have accelerated in recent years. In comparison to the Norwegian market it is in fact much larger, supporting the experts at Barclays and the ESRB conclusion of a high risk in the Swedish housing market. It must be mentioned here as well that interest rates are historically low in Sweden, as in Norway, and the fundamental factors are without doubt affected by this. However, the expected capital gain in Sweden is lower than in Norway, which in turn make it more attractive to rent than buy. This is not reflected in the graph, further implying that there could be bubble tendencies in the housing market.

51 Figure 5.13

Fundamental vs. Real P/R ratio - Sweden 6,0

5,0

4,0

3,0

2,0

1,0

0,0 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

Fundamental P/R ratio (1996 = 1) Real P/R, (1996 = 1)

Source: SCB, 2015; Svensk Mäklarstatistik, 2017; OECD, 2006; Own calculations

From our results, we have observed that the Norwegian housing market is currently showing some bubble tendencies, in the sense that the real P/R ratio has for some time now been above the fundamental P/R ratio. Also, the trend of the housing price index compared to the CPI and rent prices is significant. However, compared to its fellow Scandinavian country Sweden, the Norwegian market seems to be more stable. The Swedish housing market is showing an increasing rate of the real P/R ratio relative to its fundamental P/R ratio, which is unsettling. From this investigation, we would support the ERCB’s conclusion regarding the Swedish housing market.

5.3. Tobin’s Q In order to determine the rate of investment, economist James Tobin presented in 1969 an alternative to the neoclassical approach (Tobin, 1969). Neoclassical investment theory states that rational investors continue to invest as long as their net present value remains positive. Using the replacement cost of an investment asset, Tobin expresses the rate of investment as a function of Q, where Q is the ratio of the market value of a new additional investment asset to its replacement cost, as in equation 5.9:

52 Equation 5.9 ������ ����� �����′� � = ����������� ����

This theory therefore states that housing prices and construction costs will converge in the long run. With house prices being higher than construction costs, including the cost of construction and the cost of vacant land, developers would continue construction until prices equal each other and vice versa (Lerbs, 2012). The Q-theory is therefore useful in order to test whether market prices on housing have fundamental support from construction costs.

5.3.1. Marginal and Average Q Since the Q described by Tobin is the ratio of one additional investment asset to its replacement cost, this is the marginal Q. The one used in empirical studies is however the average Q, as the marginal Q is not directly observable. Average Q is the ratio of the market value of existing capital to its replacement cost. Fumio Hayashi (1982) derives the relationship between them, and describes the specific scenario where average Q = marginal Q:

• The suppliers in the market are price takers; The market is characterized by perfect competition, so the individual supplier has no impact on prices, and adjust quantum according to demand. In a market with price-making suppliers, average Q would be higher than marginal Q, as the market value of an additional unit of capital would be higher than the value of an existing unit of capital. • Perfect capital markets; Free flow of capital between borders is a necessary condition, as constraints on capital can limit firms from following their optimal investment strategies. • The production and installation function are linearly homogeneous with constant returns to scale: This means that an increase in input variables will result in an equal increase in the output variables.

53 Investors are more willing to invest when the value of Q is high. As more and more investors start to invest, the value of the marginal Q will decrease. As long as Q > 1, rational investors will keep investing until marginal Q = 1. Therefore, the long run equilibrium for marginal Q is the same as the long run equilibrium for average Q: Q = 1.

Figure 5.14 MCC = Marginal Cost of Capital

MPK = Marginal Product of Capital

Q>1 then MPK>MCC, so it will be optimal to increase investment

Q<1 then MPK

Source: Own creation

5.3.2. Tobin’s Q and the Housing Market The market price of housing is easily observable in the market. Comparing this to the replacement cost which is the total sum of all costs related to the construction of new housing gives us a Q-value that can help determine if there exists a bubble in the housing market. As housing varies greatly in size the market prices and replacement costs are usually measured in price per square meter.

For Q-values greater than 1, there is a profit to be made for investors with capital to invest. For the housing market, this means that home builders will be encouraged to turn uninstalled capital (lumber, nails, labor) into installed capital (Foote, 2010). Arguments can be made that a bubble exists if the market value greatly differs from the fundamental value, which is the same as market value being greater

54 than the cost of construction. Recalling that the long-term equilibrium is Q=1, we conclude that there are bubble tendencies if we see Q>1 for a longer period of time.

5.3.3. Limitations Corgel (1997) raises several concerns about using Q-theory to analyze the housing market: • The observed transaction prices might not be a good reflection of asset values. • When the ratio is used to assess development opportunities, the ratio should be assembled from the value of newly constructed assets to development cost. • The denominator should contain an adjustment to replacement cost for economic depreciation when the Q-ratio is used to evaluate price appreciation potential in the market. Newer housing will last longer and have a higher standard than old housing. • Finally, the reported ratio yields little information about the intangible values of real estate and real estate firms.

As pointed out by Corgel, newer housing will last longer and have a higher value than old housing. But including economic depreciation for old housing will be practically impossible. Buildings vary in age, standard and some have been refurbished recently while others can be virtually the same as they were decades ago. As an investor, the Q does not tell whether it is more profitable to build or buy, as it compares buying an old dwelling to the price of building a new one. Investors should compare the selling price they can get for a new dwelling to the cost of building a new if they are to use the Q-theory for investment decisions. The original Q-theory calculated the market price of a company by multiplying the number of outstanding shares with the market price of shares. This can be troublesome, as this market price contains a lot of expectations about the future. This is also true for housing prices, and therefore a variable that contains information about expectations for the future is included in later analyses. Because we add the site cost, our results might be skewed downwards. The site cost in the cities are much higher than the site cost in more remote areas in the north of Norway, thus pushing the average cost upwards. This might give a lower average Q, telling investors not to invest even though it might be profitable in some areas. This also stems from the fact that we calculate using averages for the whole country, but Norway is as previously stated a country with huge local differences in prices, supply and demand.

55 Several market frictions that may imply a slow adjustment of real residential investment to house price and construction cost movements also makes the Q-theory a less good fit for the housing market (European Central Bank, 2013): • Lack of transparency and land available for construction • Time-consuming institutional procedures for granting building permits • The time between starting and finishing a new building

This indicates that there exists a severe time-lag from planned to finished housing, thus the Q-theory might not be perfectly applicable to the housing market. Hence, observing a Q-value which is not at its equilibrium might be because of the delayed response from a change in prices or costs.

5.3.4. Data The data used to estimate prices for Norwegian houses per square meter is collected from two different sources, as they together provide a longer time series. NEF, the Norwegian Real Estate Agents Association has published prices for 1985 – 2013, while SSB has published data from 2002-2016 (NEF, 2013) (SSB, 2017g). Using two different sources for house prices could lead to problems, but the average differences during the overlapping years is only about 1%, thus supporting our decision to use two different sources. NEF is no longer the publisher of the price data, as Eiendom Norge now is the main publisher of real-estate statistics and reports. This is however a subscription-only site, and as we do not have access to Eiendom Norge’s data, we will use a tool called WebArchive in order to access earlier versions of NEF’s site. This way, we can collect the data previously published by NEF. The price per square meter published by NEF is the average square meter price for the average Norwegian 100 square meter dwelling. For the data published by SSB, we use full-owner as a proxy for the average square meter price. This is justified by the fact that over 85% of all dwellings are full-owner dwellings.

The Norwegian State Housing Bank is the Norwegian state’s central organ for the government’s housing policy. Data on construction cost is found through annually published reports by Husbanken (Husbanken, 1985-2015). Their data is based on information from approved applications for mortgages and subsidies associated with projects related to construction, both new housing and repairs on existing

56 ones. As the data is gathered at the beginning of the construction period, the actual data might differ from the published data. Included in the data is both the cost and the size of land. The cost of construction includes both labor, materials, commission and construction loans. We are able to calculate construction cost per square meter both with and without the cost of land, as well as the isolated cost of land. Some years in the reports contain few observations, weakening the validity of the data. This needs to be taken into consideration while analyzing the data.

5.3.5. Empirical Testing As stated earlier, the long-term equilibrium of Tobin’s Q is Q = 1. When Q is in equilibrium, the price per square meter for pre-owned housing is the same as the cost of construction per square meter, including the cost of land.

From figure 5.15 it is clear that for the data collected, it seems that Tobin’s Q for the Norwegian housing market actually does center around the equilibrium value, even though there are large and numerous deviations. The Q-value has fluctuated from 0.72 in the all-time low in 1992 to the all-time high in 2000 where Q equaled 1.19. Viewing these values in context with the historic development of Norwegian house prices, we see values below 1 during the Banking Crisis of the late 80’s and early 90’s, and the bursting bubble in 1987 is especially visible. The all-time low came in 1992, the same year as the real housing prices hit their lowest point. From 1988 until 1995 the housing market experienced a Q below 1, but from 1995 and onwards Q has been above equilibrium most of the time. There are however no clear indications of a bubble, as it fluctuates between 1.1 and 1 with an average of 1.04 the last 10 years, something that indicates that the increase in house prices are back by the fundamental construction cost. Figure 5.15 indicates that the housing market has been more balanced in terms of construction prices compared to house prices for the past 10 years than for the period from 1985 until 2006. There is less volatility in the Q-measure the past 10 years.

57 Figure 5.15

Tobin's Q for Norway 1,3

1,2

1,1

1

0,9

0,8

0,7

0,6

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: NEF; SSB 2017g; Husbanken 1985-2015; Own calculations

One of the arguments against using the Q-theory to analyze the housing market, is that the housing market is inelastic in the short-run, and that it takes time for suppliers to adjust to changes in demand. This can be some of the reason for the constant change in Q, and that it does not stay in the theoretical long-term equilibrium. Building dwellings is a time-consuming process, and this has caused a surplus in housing demand. In big cities, the lack of supply is intensified by the scarcity of vacant land.

The Q-value is determined by the price, cost and by the cost of land. Only analyzing the value of Q will not tell us what is driving the development in Q. To analyze whether prices or costs are the biggest driver behind the changes, they are both displayed in figure 5.16:

58 Figure 5.16

35000 House prices and construction cost, nominal per m2

30000

25000

20000

15000

10000

5000

0 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Nominal prices Nominal costs

Source: NEF; SSB 2017g; Husbanken 1985-2015; Own calculations

Comparing prices and costs, it is clear that it was the fall in housing prices that caused the value of Q to be consistently low from 1987 and until 1992. The Banking Crisis made housing prices drop, and a rapidly decreasing oil price further intensified the economic downturn of this period. Costs were at the same time relatively stable, and from figure 5.16 it seems like costs follow prices with a short time lag. An increase in housing prices is shortly followed by an increase in construction costs, and a decrease in housing prices is shortly followed by a decrease in construction cost. This highlights the lag between supply and demand and that a delayed reaction to a change in one of the variables could be one of the reasons for Q never staying at its long-term equilibrium. Throughout the last half of the 90’s, housing prices increased rapidly while construction costs took some time to catch up. The caused an artificially high Q-value that led to the drop in Q when costs started increasing following the initial increase in housing prices.

5.3.6. Comparison - Denmark It proved more challenging to find data for Denmark than for Norway. Using the construction cost index and the housing price index from DST, we compute a Q-value based on index numbers, something that only tells us the development and not the actual level of Q (DST, 2016) (DST, 2017). This makes

59 comparisons with the Q calculated for Norway less relevant, but differences in development are still possible to detect. Because of limitations in the time series published by DST, we had to limit our analysis to 2003-2015. Still, this period incorporates the boom seen in the years before the Financial Crisis and the large downturn which followed. The Q-value for the Danish market is presented below in figure 5.17:

Figure 5.17 Index Q, Denmark 1,4

1,2

1

0,8

0,6

0,4

0,2

0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: DST (2016;2017); Own calculations

The development does not seem to be as volatile as for the Norwegian market, though we are analyzing a shorter period. We also see a negative trend in Q from 2006 all the way until 2012. This indicates that the rise in building cost has been bigger than the rise in housing prices, which is also backed by an article published by Cembrit, a Danish manufacturing firm (Cembrit, 2015). The development was quite different than what was seen in Norway during that same period. In Norway, prices exceeded costs for most of the period. Like Norway, housing prices increased prior to the Financial Crisis, resulting in Q > 1 in Norway and a rising Q for Denmark in the years leading up to the crisis. When it started, the Financial Crisis hit Denmark’s housing market hard and this is clearly reflected in the figure. Prices on housing first started to recover in 2012, which is also shown in the figure. The base year here is 2015, so naturally Q is in equilibrium in 2015. However, this does not mean that the actual value of Q is in equilibrium due to the use of indexes.

60 5.4. Conclusion After conducting an empirical comparative analysis of the Norwegian housing market using the HP filter, P/R ratio and Tobin’s Q it becomes evident that the situation in Norway is more stable than the situation seen in its fellow Scandinavian countries. The HP filter showed that Norway in general is not above its long-term average, whereas both Sweden and Denmark showed to be above. However, Oslo, as the two other Scandinavian capitals, showed signs of being above its long-term average. The development of real and fundamental P/R ratios for both Norway and Sweden showed signs of bubbles tendencies. Still, Sweden seemed to be much more vulnerable than the Norwegian market, given the ratio levels. Measuring Tobin’s Q for Norway shows little signs of there being a bubble, though it is not in its long- term equilibrium. The development in Denmark shows that building costs have exceeded prices, mostly because the Danish housing market suffered hardest of all the Scandinavian countries from the Financial Crisis. To summarize, the current housing price level in Norway seems to be high, however not abnormal compared to the other Scandinavian countries.

61 6. House Price Models 6.1. MODAG Model of Aggregated Type, in short MODAG, is a macroeconomic model for the Norwegian economy, developed by Statistics Norway. It is used by SSB and the Ministry of Finance for short to medium-term analyses and forecasting of the development of the Norwegian economy. MODAG includes many variables and subsets of models. We will focus on the chapter that seeks to determine the prices of pre- owned owner-occupied housing and housing investments, and which models supply and demand for housing capital and thus housing prices (Boug & Dyvi, 2008).

Demand for housing is, according to the model, determined by disposable income, after-tax real interest rate and total housing capital. Supply is given by the existing housing stock. Changes in housing capital are determined by the aggregated supply, while prices for pre-owned housing is based on demand.

Owner-occupied housing, adjusted for inflation and new construction, can be described by partial shifts in the following explanatory factors:

Equation 6.1

��� − �� = �� + ��,� ∗ (�� − ��) + ��,� ∗ ��� + ��,� ∗ �83

The development of new builds can be described as:

Equation 6.2

���������������� = �� + ��,��� ∗ (��� − ��) + ��,�� ∗ (����83 − ��)

Where: ��� = Price on pre-owned owner-occupied housing �� = Households disposable income

��� = After-tax real interest rate

�83 = Total housing capital in constant prices

62 ����������������= Commissioning of new construction

����83 = Price index for new construction, excluding site cost

The long-term equilibrium for the implemented relation for the price of pre-owned owner-occupied housing is as follows:

Equation 6.3

��� − �� = �������� − 0.62 ∗ �83 + 1.62 ∗ (�� − ��) − 11.59 ∗ ���

For the long-term equilibrium, we see that only after-tax real interest (���), total housing capital (�83) and real disposable income (�� − ��) matters for the price of pre-owned housing. If total housing capital increases by one percent, it will reduce housing prices by 0.62 percent. If real disposable income increases by one percent, it will increase housing prices by 1.62 percent. It is also clear that the after-tax real interest has the greatest effect on pre-owned housing prices; a one percent increase in RRT results in a 11.59 percent drop in pre-owned housing prices.

The long-term equilibrium for the development of new builds is as follows:

Equation 6.4

���������������� = �������� + (��� − ����83)

So, for a given cost level, a one percent increase in pre-owned housing prices will result in a one percent increase in the development of new housing. This means that if both housing prices and building costs percentagewise increase by the same amount, the long-term development will be unchanged. Profitability is only affected by the relation between pre-owned housing prices and building costs.

6.2. Jacobsen and Naug Jacobsen and Naug published the article "What drives house prices?" in 2004, in an effort to test and detect the most important explanatory factors behind the housing prices in Norway. They estimated an

63 econometric model for the Norwegian housing market based on quarterly data from 1990 to 2004. Their goal was to explain the massive price increase from 1992 to 2004 where prices almost tripled, and to test if the market prices could be explained by the following fundamental factors: • households' total (nominal) wage income • indices for house rent paid and total house rent in the consumer price index (CPI) • other parts of the CPI adjusted for taxes and excluding energy products (CPI-ATE) • various measures for the real after tax-interest rate • the housing stock (as measured in the national accounts) • the unemployment rate (registered unemployment) • backdated rise in house prices • household debt • the total population • the shares of the population aged 20-24 and 25-39 • various measures of relocation/centralization • TNS Gallup's indicator for the households' expectations for their own and the country's economy

Creating one single house price equation that incorporated all the explanatory variables, and that gave a meaningful result, was not feasible due to the large number of factors. Instead they estimated several models that incorporated only a subset of all the variables. These models were then simplified by adding restrictions that were not rejected by the data, which made the interpretation of the dynamics easier.

Rent for housing and other consumer prices generally had insignificant effects on housing prices. The insignificant effect of rent for housing is explained by the fact that the rent in housing cooperatives accounted for a large portion of the housing rent indexes in the CPI for most of the estimation period. Also, rent has been highly regulated during the period.

The effect of the banks’ lending rate was highly significant in all the models, while the market rate was clearly insignificant on housing prices in the models in which the banks' lending rates were also included. The insignificant effect of market rates might reflect that the key rate was used to stabilize the short-

64 term development in the Krone exchange rate during most parts of the 1990's. The households might have used the observed rate as a prediction for future rate, more than they do today. The market rates may also, to a certain degree, incorporate a change in the economic outlook. It is therefore reason to believe that the effect of interest rate expectations might be undervalued in the estimated equations.

No significant effects were found on the households’ debt level on housing prices, neither when the debt variable was included in the entire period nor when it was only included for the period during the bank crisis of 1990-1993. In isolation, this implies that credit for household purchases of real estate was not limited by the profitability of banks' during the estimation period. On the other hand, there is reason to believe that other debt that the households have was affected by their profitability in the period 1990- 1993.

There was no evidence that relocation or demographic relations have a strong direct effect on real estate prices as a whole. Indirectly, demographic changes affect prices by changing the personal income in the economy. Personal income had a significant effect on real estate prices, and was included in the final model. As demographic changes happen slowly, it can be hard to capture these changes over a relatively short time span.

Households expectations about the future is largely correlated with the interest rate level and the unemployment rate. These expectations might also shift due to changes in Norway's economic outlook, a change in political conditions or negative shocks such as war, terror or a fall in stock markets. To capture the effect of these expectations an indicator for households’ expectations about their own and the country's economy is included. This indicator is a version of the expectations indicator by TNS Gallup, corrected for the effects of the interest rate level and the unemployment rate. Jacobsen and Naug first estimated a model of the consumer confidence indicator with the interest rate and unemployment as explanatory variables. Next, they calculated the difference between the actual and the fitted value of the consumer confidence indicator for each period. The difference measures changes in expectations that is caused by other factors than the interest rate level and the unemployment rate. Exactly how this is done is further explained in section 10.1.

65 The analysis done by Jacobsen and Naug concludes that the rise in housing prices over the estimation period can be largely attributed to changes in fundamentals. The expectations variable can also be used to capture the effect of non-fundamental factors, but no evidence was found that expectations had contributed to pushing housing prices up during the period. Instead they found that the rise in prices can be attributed to changes in fundamentals such as housing construction, income, interest rates and unemployment rate.

Testing various variables with quarterly data over the period they estimated a model by the method of least squares, using the factors that had the biggest effect on house prices in the estimation period. By using the nominal interest rate the model got a better fit, therefore the model expresses a connection between nominal sizes and other variables. The model Jacobsen and Naug ended with was this:

Equation 6.5 ∆ℎ��������� = 0.12 ∆������ − 3.16 ∆(�������� ∗ (1 − �)) � � � − 1.47 ∆(�������� ∗ (1 − �)) + 0.04 ����� �−1 � − 0.12 [ℎ��������� + 4.47 ∗ (�������� ∗ (1 − �)) + 0.45 ������������ �−1 �−1 �

− 1.66 (������ − ℎ�����������)�−1] + 0.56 + 0.04 �1 + 0.02 �2 + 0.01 �3

Having an explained variation of 0.8773, the model describes 87.73% of the variation in housing prices experienced in the period.

6.2.1. Weaknesses and Discussion Even though the model proposed by Jacobsen and Naug is considered a good model for analyzing the forces behind the Norwegian housing prices, there is still uncertainty about its ability to predict future housing prices. Using a model that is based on data from 1990 to the first quarter of 2004 to predict future prices could lead to the wrong conclusions, as the economic environment has changed.

66 The price increase has been most extreme in the big cities. A variable or an indicator for price differences based on urbanization should therefore have been included in the model. It is likely that urbanization has a major effect on housing prices and that it makes for even stronger bubble tendencies in urbanized areas. For these reasons, the model might not be able to identify bubbles in the housing market.

The economic situation has changed since the model was estimated. Before the Financial Crisis, the economy was more stable and there were no major banking crises’ during the estimation period used by Jacobsen and Naug. A way of including effects due to changes in the economic environment is by adding a variable that contains expectations about the future. By using the indicator estimated by TNS Gallup, Jacobsen and Naug most likely capture the effect from households changed expectations for the economy. Their estimation period also starts at the very end of a bursting housing bubble.

As this model is one of the most cited and well-known house price models for the Norwegian housing market, we will try to re-estimate it and see how well it can describe the development seen since 2004.

67 7. Fundamental Analysis of Supply Side To be able to better understand the price development in the housing market it is important to account for all price drivers. In the following section, we will conduct a fundamental analysis of different drivers on the supply side. Housing prices in the long-run should be reflected in the cost of building. Therefore, we have chosen to focus on the amount of new builds in the market, building costs, and land costs. Also, we will look at the regulations that set the framework for the banks’ lending policies. A low supply of housing will, as previously mentioned, lead to a price increase. Contrarily, a high supply will lift the pressure off the market and decrease housing prices. All numbers will be analyzed on a national level as not all factors have area specific data.

7.1. New builds The amount of new builds will greatly affect the supply side in the market. If there is an increase of new builds the pressure on pre-owned housing will ultimately decrease. It is a common opinion among experts that for some time now there has been built too few houses in Norway. In the last ten years there has been built around 250,000 new apartments in Norway. This is close to a historical low. Whereas in the 1970’s there was built 340,000 new apartments in Norway (Dagens Næringsliv, 2014). Figure 7.1 shows the historical development of new dwellings initiated each year. As one can see from the figure, 2016 was the first time since 1982 we surpassed 35,000 dwellings started, which is quite remarkable given the growth rate in the population in the same time period.

68 Figure 7.1

Dwellings Started 40 000 35 000 30 000 25 000 20 000 15 000 10 000 5 000 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Dwellings Started

Source: SSB, 2017b

New builds also vary with regards to urbanization. In urban areas the new build rate is low compared to the need. For example, in Oslo last year the city council approved the building of 5,600 new builds, whereof 1,500 was student housing. In other words, 4,100 where for non-students (E24, 2016). However, leader for the Norwegian Home Builders Association (NHBA), Per Jæger, suggested that to balance the need for housing there should be built close to 7,000 new dwellings each year in Oslo. Figure 7.2 shows SSB’s estimates and actual numbers of the need for housing and actual new builds started in Oslo. From this we can assume that there will be a long-term pressure on house prices in Oslo at least in the foreseeable future. Green columns show the estimate of need for new builds, blue columns show the actual amount of new builds started, and red columns show the amount of building approvals given by the city council.

69 Figure 7.2

Source: E24, 2016 (Prognosesenteret/Boligprodusentenes Forening)

Figure 7.3 shows the amount of new builds started and population growth each year. The number of new builds is naturally very sensitive to the startup of large single standing projects and is thus subject to large variations from year to year. From figure 7.3 we can see that since the Financial Crisis there has been built far too few dwellings relative to the need. During and after the Financial Crisis, the new build “dry up” was mainly because of low expectations in the market and high interest rates. Since there was a lower supply in the market as well as lower demand, this only resulted in a minor housing price drop. Luckily, the Norwegian housing market has a “safety-button” to press such that the process of new builds can quickly halt and all sales completed are voided, if the market should turn. In the short-run this could potentially prevent a large price drop.

70 Figure 7.3

New builds and population growth 40 000 1,40% 35 000 1,20% 30 000 1,00% 25 000 0,80% 20 000 0,60% 15 000 NEW BUILDS NEW 10 000 0,40% 5 000 0,20% POPULATION GROWTH RATE GROWTHPOPULATION 0 0,00% 2000 2002 2004 2006 2008 2010 2012 2014 2016

New builds Population growth rate

Source: SSB, 2017b; SSB, 2017f; Own calculations

When the market turned again around 2009, a low supply and a higher demand pushed housing prices upward. This pressure has only continued. Although the population growth rate has decreased in recent years, there is clearly a backlog in the need of housing which will take many years to fulfill.

7.2. Building costs The building costs are crucial for the decision of starting housing projects, and ultimately also for the supply of housing in the long-run. In theory, an increase of 1% in housing prices should in the long-run lead to an increase of new builds of 1%. Alternatively, an increase of building costs of 1% will lead to a decrease of new builds at 1% at the same price (SSB, 2011). In other words, higher costs related to building will reduce the development of new builds and lower costs will provide incentives to invest in the construction of new builds. Also, as the market for new builds and pre-owned housing are complementary, a growth in building costs should be reflected in the price of pre-owned dwellings.

71 Figure 7.4

Building cost index vs. House price index (1980 = 100) 1600 1400 1200 1000 800 600 400 200 0 1990 1993 1996 1999 2002 2005 2008 2011 2014

Building cost index House price index

Source: SSB, 2017c; NCB, 2017

As an illustration of the development between the building cost and housing prices we have collected two indexes for both factors in figure 7.4 above. We can see that building costs have steadily increased the last decades, relative to housing prices which have increased drastically. However, it is well known within the real estate development industry that SSB’s building cost index, which is used in figure 7.4, does not fairly portray the cost level these days, as it does not account for all new “cost variables” (NRK, 2013). This is mainly because of the TEK10 regulation which was introduced in 2010 (Lovdata, 2017). The index is included only to give a short-term picture of the price movements relative to each other.

The new TEK10 regulation has been a hot topic in Norway the last years with regards to how quality requirements have had an effect on the building costs and ultimately housing prices. Government requirements regarding the quality of new builds have only increased. These new regulations focus on accessibility to housing, energy efficiency, fire protection and electrical installments. All these regulations give higher building costs as they require higher standards for materials and building time for projects are longer. Since the beginning of the 20th century the productivity within building and construction has fallen (Dagens Næringsliv, 2014). This has led to a doubling of salary costs per produced housing unit, and ultimately higher building costs. Those who have to pick up most of this bill are the end consumers (NRK, 2013).

72 Although the building cost index has its flaws and is most likely set at a much lower rate than in reality, the gap between house price index and the building cost index is so large that it can only partially be explained by the building cost growth. Building costs have without doubt contributed to raising house prices in recent years. However, from figure 7.4 the gap clearly started well before the TEK10 regulation was introduced, implying that the housing price increase has not been reflected through building costs. Another aspect worth mentioning here is the land scarcity in larger cities. In these areas housing prices will be less affected by building costs, but more so for land costs.

The uncertainty between housing prices and costs becomes even more evident when we compare figure 7.4 and figure 5.16 from section 5. Whereas the cost index in figure 7.4 seems to be far below the housing price index, the costs per square meter follows housing prices per square meter relatively close in figure 5.16. As the cost index used in figure 7.4 does not include all cost variables, there is reason to believe that the cost per square meter used in figure 5.16 portrays a more accurate picture of the situation. Figure 5.16 includes all costs related to building, including the cost of land. Nonetheless, given the deviation between these two figures, one must interpret these numbers with caution.

7.3. Cost of land The cost of land in Norway has experienced similar growth rates as housing prices since 2005. By looking at figure 7.5 we can see the development between cost of land, housing price index and CPI. This is to show that the price increase for housing and the cost of land is not just from a general price increase.

73 Figure 7.5

Cost of land index, Houseprice index, CPI (2000 = 100) 300

250

200

150

100

50

0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

House price index Cost of land index CPI

Source: NCB, 2017; Husbanken, 2017; SSB, 2017; Own calculations

Available land and the amount of new builds in the market are important for the supply of land. Norway in general is a country with vast amounts of land available to build on. Statistically only two percent of the available land in Norway has been developed for housing (SSB, 2014). With that said, in urban areas such as Oslo, there is a shortage of unregulated land to build on. Politicians in Norway have great influence on the price of land through their regulation policies. Therefore, the restrictions set by the politicians on where it is allowed to build, can greatly influence the price of housing. The amount of land to build on is not the problem. The problem is that not enough pieces of land pass through the regulations set by the city councils. If more land was regulated for building, the cost of land would go down and more building projects would be released into the market, stabilizing housing prices.

Both CEO’s for two of the largest real estate developers in Oslo and Norway, Daniel Siraj with OBOS and Baard Schumann with Selvaag Bolig, agree there are too few regulated pieces of land to build on in Oslo. The capital of Norway needs more fully regulated land where the building can start earlier, instead of using two-three years to fully regulate a piece of land. As a consequence of the imbalance between supply and demand in Oslo there has been a major price increase for housing in Oslo, and most likely this will continue unless more pieces of land are regulated and regulated quicker (Dagens Næringsliv, 2015a).

74 7.4. Bank Regulations Banks are the most important source of financing for mortgages. Hence, the lending policy for banks are very important for the development of the housing market. According to the Financial Supervisory Authority of Norway, mortgages account for 60 percent of the banks total loans (FSANa, 2016). Therefore, the housing market is important for the banks’ earnings and capital coverage as well. Each bank can through the loan-to-value ratio, maturity, interest rate margin, and payment profile affect the costs of servicing the mortgage, and thereby a household’s opportunity for buying a dwelling. In Norway, mortgages have traditionally been given based on a combination between servicing ability and the collateral each household can put up as security.

7.4.1. Collateral in real estate The borrower’s access to credit is largely dependent on the value of the collateral he can put up as security, which is usually in real estate itself. This implies that the housing market has an endogenous credit rationing. Expectations of higher housing prices increase the credit availability. Simultaneously, a higher accessibility to credit will increase the demand for housing and ultimately housing prices. The opposite effect is true if there are expectations of lower housing prices. A too large emphasis on collateral in real estate can give expectation effects and credit driven housing prices, which is not desirable.

7.4.2. Loan-to-value ratio In FSAN’s circular 29/2011, new guidelines for prudent lending policies for financial institutions were published. The new guidelines were meant to halt the strong growth in the common household’s debt burden. The ideal loan-to-value ratio was therefore adjusted down from 90% to 85% (FSAN, 2011a). From FSAN’s “Boliglånsundersøkelse” in 2015 it is concluded that the percentage of households with a loan-to-value ratio above 85% had gone down to 16% (FSANa, 2016). As a comparison, in 2010, 20.7% of mortgages were above the ideal loan-to-value ratio of 90% (FSAN, 2011b). Clearly, the stricter equity requirements have had a significant impact on the debt burden for the common household.

75 7.4.3. Interest rate increases and Interest-only mortgages The new guidelines set by FSAN in 2011 also entailed the banks to account for that households must be able to withstand an interest rate increase of five percent, when reviewing their servicing ability. Also, they gave a recommendation to the banks that loans which have a loan-to-value ratio over 70% should not be granted interest-only mortgages, even for smaller periods. This is as mentioned only a recommendation and banks are in their full right to access this possibility with each customer who asks for a mortgage. Each bank has a “flexibility quota” where 10% of the mortgages they grant can void this regulation.

Both the amount of total interest-only mortgages and partial interest-only mortgages have gone down since the inception of the new guidelines. From the implementation in 2011 to 2015 the percentage interest-only mortgages went down from 23% to 11% (FSAN, 2016a). However, the average interest- only period has remained the same at 4 years (FSAN, 2012). Interest-only mortgages provide a lower liquidity strain for households, which can lead to a situation where a family takes up a larger mortgage than they otherwise would have. The banks’ stricter lending policy contributes therefore to lowering the mortgage sizes of each household.

In June 2016 FSAN suggested new guidelines as requirements for mortgages. First, the FSAN wishes to remove the banks’ possibility to void the interest-only requirements, fully removing the “flexibility quota”. Second, the FSAN suggested that each mortgage cannot surpass five times gross income. Lastly, the maximum loan-to-value amount for credit loans is reduced from 70% to 60% (FSAN, 2016b). Credit loans are flexible loans where the borrower can determine how much credit they want to use and when the installments are to start. The collateral is in the dwelling itself, and the only interest paid is for the used credit. Also, the FSAN suggested that if the Ministry of Finance in Norway does not agree to fully remove the “flexibility quota”, that they at least lower it from ten to four percent. In 2015, it was suggested that the banks increase the “stress test” limit from five percent to six percent when reviewing each household’s servicing ability (FSAN, 2015).

The average household’s debt continues to increase compared to income, which ultimately will have large negative consequences if the market drops. These new regulations help to reduce the risk of

76 households taking up more debt than they can fully service. At the same time, these regulations will result in stricter credit reviews from the banks, and thus constrain each household’s possibility to invest in housing.

7.5. Conclusion After looking at the supply side factors such as new builds, building costs, cost of land and banking regulations we can see a common trend of reasons for a housing price increase. New builds have increased in recent years; however, the number has only just reached 1980’s levels. This has created a backlog in terms of what has been needed of housing. Since the real estate development process takes years to complete it is safe to say that there will be a high demand for new builds in the years to come. This is supported by the leader for the Norwegian Home Builders Association, who said that in the last 15 years there has been only been built half of what is needed in Oslo to meet the demand (E24, 2016).

What is also interesting to observe is the large increase of building costs the market has experienced in recent years, making the profitability in the development industry constrained. These high building costs are mainly because of the new regulations set by the government, which must be followed. With that said, the lack of fully regulated pieces of land will to a larger extent affect the start of new build projects. Although the profitability is not as high as it once was for real estate developers, they all agree that the demand to build is there. Also, the availability of fully regulated pieces of land in urban areas is even smaller, making the supply side even more affected in these areas. At the same time the demand for housing is at an extreme high. When the supply side cannot keep up with the demand side, it will ultimately lead to a price increase. Therefore, we choose to conclude that a low supply side can to a large degree explain the price increase we have experienced the last two decades.

77 8. Fundamental Analysis of Demand Side In the following section, we will investigate some of the price drivers on the demand side and to what extent they can explain the housing price increase. The demand side consists of many factors. Many of them are dependent on each other and can together greatly affect the housing market. All numbers will be analyzed on a national level as not all factors have area specific data.

8.1. Disposable income A household’s disposable income sets the foundation of what financing possibilities they have, which ultimately affects the housing prices. In Case and Schiller’s article they conclude that in most states in the US, income almost completely explains housing price increases and decreases (Case & Schiller, 2004). The disposable income is a direct factor of the mortgage size one can receive from the bank at a given point in time. Therefore, if the costs related to the mortgage, such as interest and installments, increase relatively to the disposable income, many people will start struggling to fulfill their economic responsibilities.

By setting 2002 as the base year for the housing price index in Norway (“Boligprisene i Norge”) and the nominal income development (“Nominell lønnsutvikling”) we are able to see the development between the these two up until 2016. This is shown in figure 8.1 below.

Figure 8.1

Source: HolbergFondene, 2017

78 From figure 8.1 and what is written above it is clear that housing prices cannot continue to deviate from disposable income over time, at least not at the rate shown in the graph. The reason is because most housing is financed by mortgages and they are serviced by paying interest and installments using the household’s disposable income. If interest and installments grow too high compared to the disposable income, there will ultimately be financial problems for the household. If this is the case then other spending must come down, or they must sell and buy something smaller. This relationship, mortgage divided by disposable income, is “mean-reverting”. Meaning, it will typically return back to its average (Larsen, 2005). However, it must be mentioned that if the key rate is maintained at a low level, as it currently is, there is room for movement and the gap between real housing prices and disposable income can stay wide without posing any immediate threat.

If one looks at the disposable income growth in a bigger picture, and not comparing it to real housing prices, one can see another side of the situation. When looking at the average Norwegian family owning a 90 square meter apartment, 18% of the disposable income is used on interest and installments (E24, 2017). This is actually below what was the case right before and during the Financial Crisis. We have to go back to the end of the 90’s to when we first had these levels. Also, the cost of living is lower compared to before. The cost of living took 60 percent of our disposable income in 1990, this number had gone down to 35% in 2015 (E24, 2017). From these numbers, the housing price increase seems more reasonable.

Conclusively, the disposable income is high, historically speaking. However, real housing prices have only continued to increase. Holding all else equal, if the disposable income increases in a country, people will afford higher mortgages and ultimately housing prices will go up. However, it is important to point out that although the real housing prices have gone up, the housing costs, i.e. interest and installments, are at a historical low in Norway, justifying in some sense the reason for high housing prices. This is no certainty though. The key rate is low now, but eventually it will start to increase. Closely connected with the disposable income is the unemployment rate. Therefore, it is natural for us to move to this factor in the next sub-section.

79 8.2. Unemployment According to Jacobsen and Naug the demand for housing will depend on the household’s expectations for own and other’s disposable income. In other words, “increased unemployment will result in expectations of lower wage growth and increased uncertainty concerning future income and ability to repay debt. This reduces the willingness to pay for owner-occupied houses” (Jacobsen & Naug, p. 32- 33, 2004). The unemployment rate is therefore usually seen as a good indicator for the current housing prices, future housing prices and ultimately the economy as a whole.

Figure 8.2 presents the relationship between the real housing prices and the unemployment rate. In a macroeconomic sense, as Jacobsen and Naug just stated, when the unemployment rate is low housing prices should in theory go up. However, by looking at the graph one can observe that the housing prices have continually increased throughout the time period, with an exception during the Financial Crises. At the same time the unemployment rate has been much more volatile. Following the oil price drop starting in mid-2014 the unemployment rate in Norway has increased. Where one would assume to see housing prices decrease as a response to this, the housing prices have continued to increase, even at a higher rate than before, as we can see from figure 8.2. Among others, chief economist in DNB Markets Øystein Dørum, stated that this can be explained by the fact that the key rate has been kept at a record low and more importantly the prognosis for the future is that it will stay at this rate for the short- to mid-horizon, as it has two and a half years later (Aftenposten, 2014). Although the unemployment rate is higher than a couple of years ago, it is still low relative to other Scandinavian countries such as Denmark (5.8%) and Sweden (7.1%) (EuroStat, 2016). Therefore, higher housing prices can be supported by a low key rate and higher disposable income, as previously written.

80 Figure 8.2

Unemployment rate vs. House price index (1972=100) 400 7,00% 350 6,00% 300 5,00% 250 4,00% 200 3,00% 150 100 2,00% HOUSE PRICE INDEX PRICEHOUSE

50 1,00% RATE UNEMPLOYMENT 0 0,00% 1990 1993 1996 1999 2002 2005 2008 2011 2014

Real House prices Unemployment Rate

Source: NCB, 2017; SSB, 2017

8.3. Interest Rate The key rate and ultimately the lending rate from banks is said to be one of the biggest factors affecting housing prices. This is at least stated and concluded in many articles, there among in Jacobsen and Naug’s article (Jacobsen & Naug, 2004). The key rate has a large effect on the decision to purchase a dwelling because of how it will affect a household’s disposable income. A high key rate will weaken the willingness to buy housing. On the other hand, a low key rate will ultimately make it more affordable to loan money and this will consequently push housing prices up. This is the current situation in Norway, where the key rate is at a historic low of 0.50% (NCBa, 2017). If we look at figure 8.3 we can see that the national bank in Norway predicts that this low rate will most likely continue for four-five years to come. This further lifts people’s expectations of continued low interest rates and again makes it even more attractive for people to invest in housing. Expectations is another fundamental factor we will address later in this section.

81 Figure 8.3

Source: NCB, 2017

The reason for why we keep on referring to the key rate and not directly to the banks’ lending rate is because they are extremely correlated. For the banks, a low key rate means a low money market rate and capital costs. Therefore, the banks have the ability to lower the lending rate to customers and remain competitive compared to other banks. Hence, we will continue to only refer to the key rate, unless stated otherwise.

By looking at figure 8.4 one can see the relationship between the real housing price index and the key rate development. The housing price index has continuously increased since 1982 until the present day, with exceptions of the Norwegian banking crisis and the Financial Crisis. Simultaneously, the key rate has steadily decreased until today, yet at a much more volatile rate. Today’s record low interest rates, and forecasts will directly stimulate the housing market, especially in Oslo and the larger cities where there is an extreme pressure from the demand side of the market. This in turn will increase the amount of mortgage people will apply for, hence pushing for an even higher price increase in the housing market.

82 Figure 8.4

Key rate vs. House price index (1982 = 100) 350 16,00% 300 14,00% 250 12,00% 10,00% 200 8,00% 150

6,00% RATE KEY 100 4,00% HOUSE PRICE INDEX PRICEHOUSE 50 2,00% 0 0,00% 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 2015

House price index Key Rate

Source: NCB, 2017; SSB, 2017

It is important to remember that the current situation with regards to the key rate in Norway, or any other country with low interest rates for that matter, is due to the fact that the economy in the country as a whole is struggling. Therefore, in an economic sense, it is not desirable for this situation to continue in the long run. Professor at BI Norwegian Business School in Oslo, Erling Steigum explains the current situation in Norway as such, “The Norwegian Central Bank has chosen to keep the key rate in Norway low to avoid lower economic growth. Also, the key rate has been lowered internationally. If Norway has a high key rate compared to other countries, the Norwegian Krone will be very strong. Currently we have a weak Norwegian Krone, which is an advantage given the recent oil price drop, and because we need to be competitive in other sectors right now” (E24a, 2016). Although NCB currently predicts a minimal increase of the key rate, one must assume that the economy will begin to rise again in the future. When this happens, the NCB will increase the key rate to maintain a stable growth and keep the inflation rate near the long-term goal of 2.5% (NCBb, 2017). It is difficult to precisely predict the outcome in the housing market of such an increase of the key rate.

8.4. Population growth With more citizens, there is naturally a larger need for housing. In Norway, net-immigration is more important than birth surplus, as it contributes to a higher growth rate (SSB 2017e). By net-immigration

83 we mean that more people are moving into the country than out. The death rate has been declining the last century, which is also reflected in the higher life expectancy and increased population.

Keeping in mind figure 7.3 from the section above, we can observe the relationship between new builds and the population growth rate. As stated above, the growth rate of the population is important for the house price development. This is also supported by chief economist at “Prognosesenteret” Kjell Senneset, “The relationship between population growth rate and new builds are important for the price development. For 20 years now there has been built too few dwellings relative to the population growth rate, and this will most likely continue. In other words, the price driving factor of this will continue.” (Aftenposten, 2014). Throughout the period the population has experienced a positive growth rate. In these same years, the population in Norway has grown with just below one million people. Which is a significant number, as we only have 5.2 million inhabitants currently (SSB, 2017e). From this we can conclude that the market consists of many more people.

Another aspect worth looking at is the fact that urban areas are experiencing a higher growth rate than rural areas. An example of this is Oslo, which has in this same time period experienced a growth rate of 40% (SSB, 2017e). The larger cities have for some time now experienced a higher demand than supply. However, we think that the total population growth in itself has limited effect on housing prices.

8.5. Demographics Urbanization is a trend that is only becoming stronger in Norway. People are moving from the rural areas to the cities. As a result, there is an extremely high demand for housing in the city centers. The five largest cities in Norway; Oslo, Bergen, Stavanger/Sandnes, Trondheim, and Drammen inhabit around one third of the total population of Norway, and this fraction is continuing to increase. As of December 6th, 2016, the amount of people in Norway living in urban areas has reached 81% (SSB, 2016d).

Currently the age group between 16-44 accounts for about 39% of the population, and it is this age group that is most likely to move around for work and education. When doing so they tend to draw to the larger cities as mentioned above (SSB, 2017e). The social aspect is equally important as work and education

84 for this age group as well, and even more so more recently. This trend has actually been termed the “caffé latte factor” by scientist Erling Røed Larsen. In other words, people are more likely to gather around urban hot-spots with a close proximity to school, work, and other social possibilities, than to care about size and design of a household (Larsen, 2005).

This can be tied into the fact that the size of households has become smaller and smaller throughout the years, in other words, fewer and fewer people are living together. Reasons for this is that family sizes are generally becoming smaller, and there is a higher debut-age for cohabitation. Also, people are single longer and more often become single at an adult age. Interestingly enough the amount of divorces has not been as low as it was in 2016 since 1990 (SSB, 2017h). This is a positive factor in terms of housing prices, as two incomes is better than one when servicing a home mortgage. Oslo actually has the largest portion of single habitants in Norway with close to 60% of the citizens living alone in the inner-city limits (Ensliges Landsforbund, 2013). From figure 8.5 we can see the development of household sizes. As we can see, households of one or two people are the most common. The trend is clear; fewer and people are living together, hence a high demand for small households.

Figure 8.5

DEVELOPMENT IN DWELLING COMPOSITION (1920-2011) 45% 40% 1 person 35% 2 persons 30% 25% 3 persons 20% 4 persons 15% 5 persons 10% 6 persons 5% 7+ persons 0% 1920 1930 1946 1950 1960 1970 1980 1990 2001 2011

Source: SSB, 2011; Own calculations

85 This trend is clearly visible in Oslo, especially where apartments below 30 square meters are going for prices far above 100,000 NOK per square meter. In May 2016, a 19 square meter apartment went for 132,000 NOK per square meter (E24b, 2016). Whereas before it was an exception that households under 30 square meters would go for over 100,000 NOK per square meter it has now become more of a rule in Oslo. This is a direct result of the increased urbanization and demand for small households in the large cities.

8.6. Housing Taxation The taxation system in Norway today with regards to housing can be considered low, compared to other capital. In Norway housing is taxed with regards to the housing capital, property taxes, and the potential gain of a sale. Also, there are tax deductions to interest expenses as well as tax-free rental income in certain situations. In this sub-section we will go through how the housing taxation works in Norway and how this can possibly affect housing prices.

8.6.1. Tax on Housing Capital As we have previously stated, a large portion of the population in Norway have most of their capital tied up in housing. The Norwegian government has always supported this. One way of doing so is to have favorable measures when accessing the taxable value of housing. In 2010, the Norwegian government determined a new way of determining the dwellings taxable value (Skatteetaten, 2017a). The new method was developed by SSB such that it is able to move with the market value. The equation includes factors such as the dwellings’ age, location, size and type (Skatteetaten, 2017c). For primary dwellings, the maximum taxable value should not surpass 30% of market value. For secondary dwellings, the maximum taxable value should not surpass 90%. If the assessed value does in fact surpass these two thresholds, you have the right to have the value reduced.

Once you have found the taxable value of your dwelling you can use this value to find your net wealth. Your net wealth is the remainder when subtracting all debt from your total fortune. Currently, if you have a net wealth below 1,480,000 NOK you are exempt from paying a wealth tax, anything over this you would have to tax 0.85%. In other words, you can have a dwelling with a market value up to 4.9

86 million NOK which is self-financed and you still do not need to pay a wealth tax1. On the other hand, having 4.9 million NOK in a bank account, would result in you having to pay 29,359 NOK in taxes2. Having these conditions clearly makes it more favorable to have your money in housing rather than in a bank account.

8.6.2. Property Taxes The property tax is a voluntary tax form decided by each municipal. In 2016, 365 of 428 municipalities in Norway had applied the property tax. Although it varies from municipality to municipality, the taxable value used to calculate the property tax is usually SSB’s equation. Using the equation from SSB, the taxable value is set to maximum 30% of the market value of the dwelling. Each municipality can choose to have a property tax rate of between 0.2% to 0.7% (Huseiernes Landsforbund, 2015). The inception of a property tax will not lead to an immediate drop in house prices as the tax rate is so small, as well as the fact that most municipalities has this tax, therefore it is almost inevitable that one has to pay it.

8.6.3. Tax Deductions on Interest Expenses The tax deduction on interest expenses is not directly aimed towards households who own a dwelling. However as previously stated, since such a large portion of the Norwegian population own their own dwelling, the tax policy is very favorable for a large portion of Norwegians. The tax is only an act of tax symmetry. As interest income is taxed at a 24% level, all interest expenses must be deducted at a 24% level (Regjeringen, 2017). Both these levels have gone down from 27% only two years ago, giving a tax break for income, but at the same time making it relatively more expensive to service your mortgage. From this one would expect to see the demand go down, however this has not been the case and the demand has only gone up.

8.6.4. Tax on Sale Profit Any sale of a dwelling in which results in a profit the owner or owners must pay a tax of 24%. However, the tax is voided if both these two statements are fulfilled.

1 (1,480,000/0.30) = 4,933,333 2 4,933,333 * 0.85% = 29,353

87 1. “The sale must take place or be agreed upon more than one year after the property was acquired, and 2. The owner must have used the property as his own home for at least one of the past two years before the sale takes place” (Skatteetaten, 2017a). Also, one will receive a tax deduction at the same rate if the sale results in a loss. The tax in itself does not primarily effect the average household, as they usually live longer than two years in the same dwelling. One can say that this tax helps keep many speculators out of the market as it lowers the profitability for short-term investments.

8.6.5. Tax on Rental Income Rental income for a dwelling owner is rent free if less than 50% of the dwelling, measured by rental value, is rented out. Also, the rental income is tax free if a larger portion or the entire dwelling is rented out and the rental income does not surpass 20,000 NOK (Skatteetaten, 2017b). These favorable tax policies give incentives for people to buy dwellings with the possibility to rent out a section. From this, these policies could lead to a higher demand for larger dwellings, which in turn would lead to a higher debt burden.

All these favorable tax policies for dwelling owners allows the potential returns after tax to be higher than the alternative cost of placing their money in other assets. This ultimately leads to a larger portion of the Norwegian population who invest their money in housing. This will again lead to higher housing prices. We think that the tax deduction for interest expenses have the largest effect on each household’s behavior towards investing in housing. Conclusively, housing taxation in Norway is clearly a large factor in the housing price development.

8.7. Expectations The housing market is largely driven by the consumer’s expectations. Expectations of future increases in housing prices provides an incentive to buy today. This will in turn lead to an increased demand. On the other hand, expectations of higher prices will also lead to people holding on to their dwelling longer to be able to benefit from potential price increases. This is something we are currently experiencing in

88 Oslo as people are waiting to sell their own dwelling before they have purchased something new. This is in fear that they will not be able to buy something new for the same sell price. This is solely made from the situation that people expect the prices to go further up. Ultimately this leads to less houses available and a higher demand for housing, hence higher prices. Daniel Kjørbeg Siraj, CEO for OBOS in Norway explains the situation like this, “clients are telling us that they are scared to sell their house before they buy something new. Simply because of the fact they do not know if they will find a new house they can afford. People have a tendency to buy themselves upward in the housing market. However, right now there are so few objects available, and because of the high price increase they risk not being able to keep up when the bidding rounds go. Also, people would rather not be back in the renting market” (Dagens Næringsliv, 2016a).

Increased expectations lead to increased activity in the market. This results in increased demand for mortgages. In a rising market the banks will also be optimistic. The effect is self-reinforcing. The banks will lighten their conditions and more people will get their mortgage approved. The banks believe that their clients will meet their payment obligations and as a result they will require a smaller risk premium. This will lead to a reduced real interest rate. The amount of money in the market will therefore increase. Conclusively this means that dwelling buyers can more easily finance the housing price increase from what the expectations of higher prices creates.

However, much of this is not true anymore in Norway, at least in terms of the banks. Because of the situation, regulations regarding approval of mortgages have increased. Now everyone must have at least 15% equity when purchasing a house in Norway, up from 10%. Also, a client cannot receive a mortgage of more than five times their gross income. More recently the Norwegian government has introduced specific regulations for the market in Oslo. The most significant is the requirement of 40% equity for the purchase of a secondary dwelling in Oslo. This is strictly to make it more difficult for speculators to buy dwelling in Oslo.

89 8.8. Conclusion In our chosen period to look at, between 1980-2017, housing prices have experienced a continuous growth. Also, looking at the factors we have written about, disposable income has increased while the key rate has decreased to record low levels. As housing prices have increased, so has the value of the collateral people have secured their mortgages with. This has resulted in an overall increase of the public debt burden. The population has experienced a continuous growth, while the unemployment rate had a steady decrease up until the oil price drop in 2014. All these mentioned factors will to some extent affect the demand of housing.

It is interesting to observe that housing prices nationally have not decreased given the higher unemployment rate. However, regionally the unemployment rate has had a negative effect, for example in Stavanger, resulting in a price decrease. Also, the urban trend, led by the “caffé latte factor”, has most likely contributed to higher prices in urban areas, compared to rural areas.

Conclusively, we think that the higher disposable income, low key rate and higher debt burden has had the largest effect on the price increase we have experienced in this period. Also, we feel that the expectations for the future in the market are pivotal for further price development.

90 9. Correlation Analysis In sections seven and eight we discussed various price drivers on both the supply side and demand side of housing. All the factors we wrote about affect housing prices more or less in some way, and there are many other smaller factors we could have mentioned as well. In the following section our goal is to look at six of the factors we have just written about and conduct a correlation analysis. We want to see to want extent these factors are correlated with the housing price development between 1990 and now. As we will be re-estimating Jacobsen and Naug’s model in the next section, we will be using both their time period of estimation, as well as after, to see how the results have changed.

9.1. Analysis To be able to investigate whether or not there is a correlation between housing prices and the various price drivers we have chosen to conduct a correlation analysis. The Pearson correlation coefficient measures the correlation between two variables (Pearson, 1895). The correlation coefficient always has a value between -1 and 1. A correlation equal to 1 entails a perfect correlation, while a correlation equal to -1 entails a perfect negative correlation. When the correlation is positive, an increase in one of the variables will always result in an increase in the other variable. For a negative correlation, an increase in one variable will result in a decrease in the other variable. If there is a correlation coefficient of 0, there is no linear connection between the variables. Also, it is important to point out that correlation does not imply causation between the two variables (Tufte, 2006).

Table 9.1 presents the correlation between six various price drivers of housing prices. As stated above we will be looking at Jacobsen and Naug’s estimation period between 1990-2004, the period after 2004- 2016, and the whole period 1990-2016. The reason for why we are dividing the periods up this way is to point out how these variables change depending on which period one is looking at. This is also important to remember for the next section when we will re-estimate the Jacobsen and Naug model.

91 Table 9.1 Time period: 1990 - 2004 2004 - 2016 1990 - 2016

Correlation with: Housing Prices Housing Prices Housing Prices New Builds 0.8438 0.2523 0.7885 Disposable Income 0.9768 0.9598 0.9882 Key Rate -0.5334 -0.5007 -0.7849 Unemployment Rate -0.6549 0.1447 -0.5210 Building Costs 0.9854 0.9722 0.9936 Population 0.9704 0.9750 0.9882 Source: SSB, 2017 (New builds, Disp. Income, Unemployment Rate, Building Costs, Population); NCB, 2017 (House price index, Key Rate); Own calculations

The correlation test shows that disposable income, building costs and population have a rather close to perfect positive correlation with house prices for the full time period. This indicates that these variables all move in the same direction as housing prices. This also supports our fundamental analysis from the two previous sections where we found that all of these variables have a positive effect on housing prices. Therefore, this development could have led to higher dwelling prices.

The variable new builds has a slightly lower correlation coefficient than the others, although it is still significant. The reason for this is most likely because of the highly volatile development of new builds. We have seen that there were very few houses built after the Financial Crisis, while housing prices kept increasing. From this, we cannot say that new builds have a perfect correlation, although we know from previous analysis new builds affect dwelling prices in the long run.

The key rate and unemployment rate both have a negative correlation and indicates that housing prices move in the opposite direction of these variables. The strongest of the two is the key rate. This also supports our fundamental analysis where we conclude that decreasing interest rates has been one of the strongest price drivers for housing prices. The unemployment rate is also negative, but not as much. This is most likely because the unemployment rate has had certain periods where it has increased while dwelling prices have kept going up.

92 For the correlations during and after Jacobsen and Naug’s estimation period the biggest changes happened to be for new builds and unemployment rate. During Jacobsen and Naug’s estimation period new builds had a positive correlation of 0.84, while in the period after until today it dropped to 0.25. As mentioned earlier, the development of new builds is very volatile. In this case, in the years after the Financial Crisis, the new builds in Norway dropped significantly, while prices kept increasing. It is important to point out that the low development of new builds was in fact to keep the price increase at a positive and stable level. What is most interesting to notice is how the correlation coefficient for unemployment rate goes from being a negative number of -0.65 to 0.14. This implies that in the time period of 2004-2016 there was a positive correlation between the unemployment rate and housing prices. In a macroeconomic setting this is very counterintuitive. However, this just points out how different certain periods can be from each other.

To summarize the correlation analysis, we can clearly see that our fundamental analysis is backed by the results for the full time period of 1990-2016. The fundamental factors have to a large extent affected the price development of housing the last 25 years. With that said, when looking at more specific periods the results can vary quite a bit. This is important to remember when we are now going to re-estimate the Jacobsen and Naug model for the time period 1990-2004.

93 10. Re-estimation of the Jacobsen and Naug model After reading and writing about Jacobsen and Naug’s housing price model from 2004 we became interested to see how well this model can explain today’s price changes. This, as well as how the coefficients have changed from 2004 to 2017. One of Jacobsen and Naug’s original model’s weaknesses was that during the whole testing period between 1990-2004 there were no major economic crises the model had to deal with. The Norwegian economy had just recovered from the Norwegian Banking Crisis and the dot-com bubble in 2000 had no significant effect on Norwegian housing prices. Now, 13 years later we have experienced a major global financial crisis, a major drop in oil prices, and the economic conditions are in general different than in 2004.

In the following section, we will present a re-estimation of Jacobsen and Naug’s house price model. From this we will discuss the various coefficient’s effect on house prices, as well as how they have changed from when Jacobsen and Naug published their findings. We will also test the model for various pitfalls through time series analysis. We used Excel to run the regression on the data, and did various testing in the statistical program R-Code. To keep consistency in the model we have chosen to use the same variables and the same approach as Jacobsen and Naug used in their original model. It was therefore very beneficial for us when the co-author Bjørn E. Naug sent us the Excel spreadsheet they used for the analysis, naturally with all the same variables (Naug, 2017). However, it was now updated with all recent numbers to date. As a disclaimer to us, Naug said that the numbers where not exactly the same as when they conducted the analysis, which would result in different numbers if we ran a regression in their time period. Nonetheless, the variables where the same and noted in the intervals we needed. The only aspect which has changed for our analysis is the length of the time series, which in our case is from quarter one in 1990 to the second quarter of 2016.

10.1. Expectation Variable In Jacobsen and Naug’s model the expectation variable is based on TNS Gallup’s indicator of households’ expectations. This trend indicator is not adjusted for seasonality or random variations. The variable that was given to us and is the same as one can download on TNS Gallups’s website as well. We are uncertain about why the authors have chosen to use unadjusted numbers for this variable. As we

94 stated above, we wish to uphold consistency when applying the model, and therefore we will use the same numbers. Also, it is stated in the article, that the variables with small letters in each equation, indicate that the variables are measured on a logarithmic scale (Jacobsen & Naug, 2004). This is an approach we have followed as well and is true for all equations with small letters throughout this section.

As we previously wrote in section 6 above, Jacobsen and Naug created a trend indicator that was adjusted for the effects of interest rates and unemployment. The reason for this was because TNS Gallup’s trend indicator is strongly correlated with the interest rate and unemployment rate, which are specified as separate explanatory factors in the model. The variable EXPEC contains therefore the expectations households have for the future, in which can be explained by other factors than interest rate and unemployment. Such factors could be changed political conditions, changed views for the Norwegian economy, and negative shocks such as war, terror and a stock market crash (Jacobsen & Naug, 2004). We have chosen to follow the same approach as Jacobsen and Naug to estimate the expectation variable. The model is found in equation 10.1 below.

Equation 10.1 ∆� = �������� − � ∆(�������� ∗ (1 − �)) − � ∆������������ − � � − � �������� � 1 � 2 � 3 �−1 4

∗ (1 − �)�−1 − �5�������������−1 + �6�1 + �7�2 + �8�3

The residual from the regression in equation 10.1 is the part of the trend indicator which cannot be explained by the interest rate or unemployment rate. The variable EXPEC is defined in equation 10.2 below.

Equation 10.2 ����� = (� − �) + 100 ∗ (� − �)3

The constant E is TNS Gallup’s original trend indicator, and F is the part of E that is explained by the interest rate and unemployment. (E – F) is therefore the part of E which cannot be explained by the interest rate and unemployment rate, in other words the residual of equation 10.1. Before (E – F), i.e. the

95 residual, is inserted in equation 10.2 the residual is added for two quarters, t and t-1. The authors do not explain why the last part of equation 10.2 is included. One explanation could be that they want the variable to be three times differentiable to capture the curve, that is the degree of concavity or convexity. Another explanation could be that they wanted to smooth out differences.

In table 10.1 we have gathered the original numbers for the expectation model from Jacobsen and Naug’s study in 2004, a re-estimate of the expectation model for the same time period using the updated data, and lastly a re-estimate of the expectation model for our time period of 1992-2016.

96 Table 10.1 Original data from Re-estimate 1992 - Re-estimate 1992 - 1992 - 2004 2004 2016 ∆�� Coefficients t-stats Coefficients t-stats Coefficients t-stats ∆(�������� ∗ (1 − �)) -12.96*** (6.68) -12.90*** (6.83) -11.99*** (6.12) �

∆������������� -0.43** (2.47) -0.44** (2.54) -0.47*** (3.43)

��−1 -0.11 (1.06) -0.13 (1.31) -0.10* (1.68)

�������� ∗ (1 − �)�−1 -0.40 (0.42) -0.65 (0.69) -0.34 (0.75)

������������ �−1 -0.03 (0.82) -0.02 (0.61) 0.004 (0.14)

0.21*** (4.57) 0.21*** (4.71) 0.20*** (5.99) �1

�2 0.10*** (4.49) 0.09*** (4.49) 0.07*** (3.82)

�3 0.22*** (5.61) 0.22*** (5.67) 0.18*** (6.19)

�������� -0.07 (0.39) -0.15 (0.91) -0.08 (0.72) Observations 46 46 95 R2 0.80 0.80 0.58

*Significant at a 10%-level, **Significant at a 5%-level, ***Significant at a 1%-level

The reason for why we added the re-estimate of the same period as the authors is to punctuate the fact that although we have the same variables and dataset, there are certain changes to the numbers which result in slightly different numbers than the original model. However, the explanatory factor of 80% is the same which provides a validity of the re-estimated model. The re-estimated model for our time period naturally has some larger deviations than their model. The explanatory factor of the model is smaller for

97 our time period. In other words, the interest rate and unemployment rate explains 58% of the variation in the trend indicator.

10.2. Re-estimation of Jacobsen and Naug model Before we present our results from our re-estimation the original model is repeated below. To make it easier for us to interpret the coefficients when applying the model, we will open the parenthesis for the error term and thus change the sign within. Other than this all coefficients are the same.

Equation 6.1 ∆ℎ��������� = 0.12 ∆������ − 3.16 ∆(�������� ∗ (1 − �)) � � � − 1.47 ∆(�������� ∗ (1 − �)) + 0.04 ����� �−1 � − 0.12 [ℎ��������� + 4.47 ∗ (�������� ∗ (1 − �)) + 0.45 ������������ �−1 �−1 �

− 1.66 (������ − ℎ�����������)�−1] + 0.56 + 0.04 �1 + 0.02 �2 + 0.01 �3

The results of our re-estimation are found in table 10.2 below. The table includes the original results from 2004, a re-estimate of the model in their time period using the updated numbers, and lastly a complete re-estimate of the model using the updated numbers and our extended time period of 1990- 2016.

98 Table 10.2 Original data from Re-estimate 1990 - Re-estimate 1990 - 1990 – 2004 2004 2016 ∆ℎ���������� Coefficients t-stats Coefficients t-stats Coefficients t-stats 0.12* (1.94) 0.17 (0.25) 0.47 (1.53) ∆�������

∆(�������� ∗ (1 − �)) -3.16*** (7.04) -3.02*** (6.07) -3.02*** (6.13) �

∆(�������� ∗ (1 − �)) -1.47*** (3.27) -1.70*** (3.46) -0.85* (1.78) �−1

������ 0.04*** (3.09) 0.03** (2.47) 0.02*** (3.10)

ℎ����������−1 -0.12*** (5.69) -0.14*** (4.63) -0.07*** (3.25)

�������� ∗ (1 − �)�−1 -4.47** (2.54) -3.60* (1.81) -13.29*** (4.94)

������������ � -0.45*** (3.48) -0.29** (2.59) -0.29* (1.71)

(������ 1.66*** (8.63) 1.83*** (4.12) 1.29* (1.89) − ℎ�����������)�−1

0.04*** (3.35) 0.04*** (6.88) 0.04*** (8.24) �1

�2 0.02* (1.80) 0.02*** (3.42) 0.02*** (4.46)

�3 0.01 (0.73) 0.00 (0.42) 0.00 (0.77)

�������� 0.56*** (3.42) 1.34*** (3.94) 0.56** (2.19) Observations 56 56 105 R2 0.88 0.84 0.74

*Significant at a 10%-level, **Significant at a 5%-level, ***Significant at a 1%-level

99 The re-estimate of the model for the same time period as Jacobsen and Naug is included for the same reason as for the expectation variable. We chose to include this re-estimate to point out the fact that although the regression is run for the same time period the variables turn out to be different, significantly in some cases. As Bjørn E. Naug noted when he sent us the spreadsheet, the numbers are not quite the same and we cannot expect to get the same numbers as they did in the original model from 2004. Therefore, these updated numbers are included as a comparison for our re-estimation for our time period. Before we comment and discuss the results of the analysis we will conduct various time series analysis to investigate whether or not the numbers are reliable.

For the coefficients inside the square brackets, that is, �������� ∗ (1 − �)�−1, �������������, and

(������ − ℎ�����������)�−1, we have divided the corresponding variables by 0.07, which is the variable outside the bracket. This is to make our variables relatable to the variables of Jacobsen and Naug. As we must open up the bracket to run the regression, we can only assume that this is the approach used by the authors as well. They have not written anything about this in the article, and therefore we choose to make this assumption to move forth with the analysis. The numbers will be interpreted as they are presented in table 10.2.

10.3. Testing the model In order to conduct valid tests on the model parameters, the residuals have to be normally distributed (Brooks, 2008). By plotting the residuals, we get a visual test for the distribution of the residuals. From figure 10.1 below, we can see that the residuals display a large degree of symmetry around a mean of about zero, which indicates that there exists a relatively high degree of normality in the residuals.

100 Figure 10.1

25 Distribution of residuals

20

15

10 Frequency 5

0 -0,04 -0,033 -0,025 -0,18 -0,01 -0,002 0,005 0,013 0,02 0,028 More

Source: Naug, 2017; Own calculations

10.3.1. Testing for Stationarity For us to be able to trust our tests, we have to test for stationarity in our data series. Stationarity means that the time series does not “wander off” over time to plus or minus infinity (Cuthbertson & Nietzsche, 2004). If the time series are non-stationary we might get spurious results from our analysis, weakening the strength of our test. We plot the variables development over time, in order to visually determine whether they exhibit stationarity or not. We will then continue with more in-depth testing of the data used in the model.

From looking at figures 10.2-10.7, it seems like total housing mass, housing prices and total income in the economy all have a positive trend. The indicator for expectations about the economy exhibits neither a positive nor a negative trend, which is a good indicator for stationarity. As it fluctuates around zero, it also does not seem to have an intercept. The after-tax interest rate, and to a certain degree the unemployment rate, exhibits negative trends. These are difficult to explain using economic theories, but it might just stem from the relatively short time-span used in our analysis. The rate in Norway is now at a historic low, and the unemployment rate has stabilized at a lower level than it was in the beginning of the period, resulting in what might seem like a negative trend. For longer time series, these variables are likely to be stationary.

101 Figure 10.2 Figure 10.3 Total housing stock Unemployment rate 3500000 7,00 6,00 3000000 5,00 2500000 4,00 3,00 2000000 2,00 1,00 1500000 0,00 1990Q2 1991Q4 1993Q2 1994Q4 1996Q2 1997Q4 1999Q2 2000Q4 2002Q2 2003Q4 2005Q2 2006Q4 2008Q2 2009Q4 2011Q2 2012Q4 2014Q2 2015Q4 1990Q2 1991Q4 1993Q2 1994Q4 1996Q2 1997Q4 1999Q2 2000Q4 2002Q2 2003Q4 2005Q2 2006Q4 2008Q2 2009Q4 2011Q2 2012Q4 2014Q2 2015Q4

Source: Naug, 2017 Source: Naug, 2017 Figure 10.4 Figure 10.5 After-tax interest Income 12 250000 10 200000 8 6 150000 4 100000 2 0 50000 1990Q2 1991Q4 1993Q2 1994Q4 1996Q2 1997Q4 1999Q2 2000Q4 2002Q2 2003Q4 2005Q2 2006Q4 2008Q2 2009Q4 2011Q2 2012Q4 2014Q2 2015Q4 1990Q2 1991Q3 1992Q4 1994Q1 1995Q2 1996Q3 1997Q4 1999Q1 2000Q2 2001Q3 2002Q4 2004Q1 2005Q2 2006Q3 2007Q4 2009Q1 2010Q2 2011Q3 2012Q4 2014Q1 2015Q2 Source: Naug, 2017 Source: Naug, 2017 Figure 10.6 Figure 10.7 House price index TNS Gallup 300 50 40 250 30 200 20 150 10 0 100 -10 50 -20 0 -30 1990Q2 1991Q4 1993Q2 1994Q4 1996Q2 1997Q4 1999Q2 2000Q4 2002Q2 2003Q4 2005Q2 2006Q4 2008Q2 2009Q4 2011Q2 2012Q4 2014Q2 2015Q4 1990Q2 1991Q4 1993Q2 1994Q4 1996Q2 1997Q4 1999Q2 2000Q4 2002Q2 2003Q4 2005Q2 2006Q4 2008Q2 2009Q4 2011Q2 2012Q4 2014Q2 2015Q4

Source: Naug, 2017 Source: Naug, 2017

102 Only doing a visual check is not enough to fully determine whether the data is stationary or not. Therefore, we perform an augmented Dickey-Fuller test to test for a unit root in the time series. A process with a unit root does not respond well to shocks, as shocks have a permanent effect, making the time series “wander off” towards plus or minus infinity. The augmented Dickey-Fuller test tests the null hypothesis of whether a unit root is present in the sample against the alternative hypothesis that the time series is stationary or trend-stationary. The test statistic used in the test is a negative number, and the more negative the number the stronger the rejection of the null-hypothesis. We use a program called R- Code to test our time-series for stationarity. R-Code has a built-in function that runs the ADF-test on any given time series, and returns the p-value and the Dickey-Fuller test statistic. The output from the tests are shown in table 10.3 below. When used on the raw data we find that only the created variable EXPEC is stationary. In the Jacobsen and Naug model, we use the first difference of the variables income, house prices and after-tax interest rate. This is a normal way to render the data stationary (Brooks, 2008). Our test is therefore done on the transformed data.

Table 10.3 EXPEC p = 0.0406 House price, first difference p < 0.01 After-tax interest, first difference p < 0.01 Income, first difference p = 0.286 Total housing stock p = 0.01546

After first differencing the variables, most are now stationary at the 10% and 5% level. The after-tax interest rate and housing prices are both statistically significantly different from H0 at the 1% level, while income still is not stationary when first differenced. When plotted, income seems to have a different development compared to the other variables. It might be more affected by changes in the economy and therefore it will follow a more cyclical pattern, being more trend-stationary than the others.

103 10.3.2. Testing for Autocorrelation Autocorrelation means that there is a pattern that is being repeated over time, and that the process is dependent on previous values of itself. Since we have a different dataset than the original used by Jacobsen and Naug, we have to test for autocorrelation. If the model contains autocorrelation, the solution will not minimize the total variance in the future and the model will not be usable for predicting future house prices. By first doing a visual test, we can determine whether the data contains positive or negative autocorrelation (Brooks, 2008). Figure 10.8 and 10.9 shows the residuals plotted both against time and against the immediately previous one, respectively.

Figure 10.8 Figure 10.9

0,04 0,04 0,03 0,03 0,02 0,02 0,01 0,01 0 0 -0,01 -0,06 -0,04 -0,02 -0,01 0 0,02 0,04

-0,02 1990Q2 1991Q4 1993Q2 1994Q4 1996Q2 1997Q4 1999Q2 2000Q4 2002Q2 2003Q4 2005Q2 2006Q4 2008Q2 2009Q4 2011Q2 2012Q4 2014Q2 2015Q4 -0,02 -0,03 -0,03 -0,04 -0,04 -0,05 -0,05

Source: Naug, 2017; Own creation Source: Naug, 2017; Own creation

There are no clear trends or patterns seen in the plotted residuals. Against the lagged values, the points are randomly spread out over all quadrants, though it might seem like there exists a weak positive trend. Plotted against time, the residuals cross the x-axis neither too frequently nor too little. These are almost exactly the results we want, as this indicates no autocorrelation. The weak positive trend in the lagged values could however pose a problem. Simply plotting the residuals is, however, not a strong enough test for autocorrelation. Therefore, we must use statistical tests to strengthen our analysis.

The Durbin-Watson test will not be a wise choice on this model, as no lagged dependent variable can be found among the regressors when using the Durbin-Watson test (Brooks, 2008). In the Jacobsen and

104 Naug model, housing prices do in fact appear in a lagged form among the regressors. Therefore, we instead we will use the Ljung-Box test to test for autocorrelation, as this allows for lagged values of the dependent variable.

The Ljung-Box test is used to test for autocorrelation. The null hypothesis is that there is no autocorrelation, and the Q-statistics is chi-square distributed with m degrees of freedom (Brooks, 2008).

Equation 10.3 � �̂2 � = �(� + 2) ∑ � ~�2 � − � � �=1

T = sample size m = maximum lag-length 2 �̂� = the estimated autocorrelation coefficient for a given number of lags, k

We use R-Code to calculate the Ljung-Box Q-value, and the output from R is seen below:

Q(1) = 6.9069, p-value = 0.008586 Q(4) = 24.631, p-value = 0.00005967

For both one and four lags, the test statistics suggest that there actually is autocorrelation in the residuals. Q is statistically significantly different from zero at the 1% confidence level for both lags. This is not the same conclusion we found in the graphical plot, and therefore it is hard to tell whether there actually is autocorrelation in the residuals. Autocorrelation is a normal problem with time series, and it goes against the requirement of OLS that the error terms need to be independent. In conclusion, the time series exhibits minor issues that could weaken the validity of the results, but not more than what could be expected of real-life time series.

105 10.4. Interpretation of the Coefficients With the conclusion of the last sub-section in mind, we can move forth with the analysis and interpret the results without altering the data. The explanatory factor in the original model was 88%, which is fairly high. For the re-estimation of the model in our time period the explanatory factor dropped down to 74%. From this we can conclude that the model does not explain all variations in housing price, 26% of the variation is explained by other variables. The coefficients of the independent variables affect housing prices in the same direction as before. However, the values of the coefficients are for the most part different.

The coefficient for ∆������� has gone up from 0.12 to 0.47. The change in housing prices will from this increase with 0.47% if the change of income increases by one percent in quarter t. ∆������� shows a significant increase. Our model shows that an increase in income in the short term will have a larger effect in house prices than Jacobsen and Naug concluded. From previous analysis in section 8 we know that a change in disposable income will have a positive effect on house prices. A household’s disposable income is essential when financing a dwelling and will to a large extent determine the mortgage one can receive. We think the significant increase in the effect of a change in income in our model gives a more accurate picture of changes in housing prices than what Jacobsen and Naug’s model did.

Further, the coefficient for the variable ∆(�������� ∗ (1 − �)) has gone slightly down from -3.16 to � -3.02. The change in itself is not notable and cannot be explained by macroeconomic changes. Our results conclude that a one percent increase in the change of interest rate in quarter t will result in decrease in the change in housing prices of 3.02% in quarter t. We know this from our fundamental analysis in section 9 where we explained the relationship between the interest rate and housing prices. Holding all else equal, a decrease in the interest rate should lead to an increase in house prices, and vice versa. This is backed by the assumption that with a lower interest rate the common household will have a higher disposable income, and thus more money to spend on housing.

Closely related to the previous variable is ∆(�������� ∗ (1 − �)) , which explains the effect on the �−1 change in housing prices by a change in interest rate in quarter t-1. Our coefficient increases a bit

106 compared to the authors coefficient, from -1.47 to -0.85. This result implies that a one percent increase in changes of interest rate in quarter t-1, will lead to a decrease in the change of housing prices of 0.85% in quarter t. In other words, this is a lagged effect of a change in interest rate. The results, both ours and the authors, surprise us. The reason for this is that one would think that an interest rate change would have a larger effect in quarter t-1, than in quarter t as it takes time for a market to react to a macroeconomic change. However, the model could be implying that housing prices are in general more responsive to interest rate changes in the short term, rather than the long term. This is something we discuss more closely later in this section.

The expectation variable has decreased to half its previous rate, from 0.04 to 0.02, and consequently affects the change in housing prices less than before. Expectations is also a factor we discussed in section 9, where we wrote about the importance of expectations in the market when it comes to housing price development. Economic theory often points to the fact that it is often expectations of future profits which stimulates demand, and therefore price growth. High expectations are often significant when the market is in a bubble. Both the authors and our results surprise us first of all because of the level of effect, and secondly that for our time period the level has gone down. If Norway has been in somewhat of a housing bubble the last 20 years, one would assume that expectations would have a larger effect on house prices.

The coefficient for ℎ����������−1 determines how the change in housing prices are affected by other factors than those included in the model. The variable has been increased from -0.12 to -0.07. This implies that the re-estimated model to a larger extent allows housing prices to deviate from, for example income growth, without meaning the housing prices are significantly over their fundamental value.

We experienced the biggest deviation from the original model with the coefficient �������� ∗ (1 −

�)�−1. The variable in the original model was -4.47 and our re-estimated model had a value of -13.29. The decrease is so significant that we choose to interpret this coefficient with caution. Our re-estimated model implies that a one percent increase in interest rate, will decrease the change in housing prices by 13.29% in quarter t-1. The average household’s debt level has grown since the beginning of the 00’s. Also, equally as drastic, the key rate has dropped in the same time period. The long-term effect of the interest rate has become stronger in recent years. Therefore, one can say that the interest is the variable

107 which has the strongest effect on house prices in the model. The results for this coefficient will be discussed in further detail later on in this section.

The coefficient for the unemployment rate seems to have a smaller effect on housing prices. In the original model if the unemployment rate increased by one percent it would lead to a reduction in the changes of housing prices by 0.45%. In our re-estimated model this value has decreased to -0.29%. One reason for why the coefficient is low is because it’s not lagged, which we find strange. Especially since this coefficient is in the square brackets of the equation, which is said to measure the long-term relationship of the coefficients and house prices. An increase in the unemployment rate will most likely not affect housing prices the same period the increase incurs. We find our re-estimated value too low, and it is unlikely that the unemployment rate will have such a low effect on housing prices. With that said, from our analysis in section 8 we saw that the rise in unemployment rate the last years has not negatively hit the housing market for Norway in general.

Lastly is the coefficient (������ − ℎ�����������)�−1. This variable has also decreased, from 1.66 to 1.29. The results imply that a one percent increase of the housing stock will decrease the change in housing prices by 1.29%. For the same percent increase in income it will generate a corresponding increase in changes of housing prices. We find this value too low as well. Too low with regards to what we find realistic to the real world. From our fundamental analysis in section 7 and 8 we know that the supply is far too low compared to the demand. From this we would expect the relationship between income and housing stock to have a larger influence on housing prices.

10.5. Discussion of model The house price model Jacobsen and Naug created in 2004 is considered to be a good explanatory model of what drives housing prices in Norway. However, how well the model works to predict future housing price movements is a topic of discussion. This is mainly because of the fact that the estimation period taken into consideration by the authors is only from the first quarter of 1990 to the first quarter of 2004.

108 Jacobsen and Naug calibrated the model to make it match the price level of the market, and therefore it will always be in equilibrium. What is interesting to notice is how the model fits the price movement of housing prices in the authors estimation period compared to our period. As one can see in figure 10.10 the model moved very close to the market during their estimation period. Figure 10.10

Jacobsen & Naug Model Vs. House price index (percent change over four quarter period) 1990-2004 25% 20% 15% 10% 5% 0% -5% -10% -15%

J&N (% Change) House price index

Source:Naug, 2017; NCB, 2017; Own calculations

If one looks at our estimation period there is a much larger deviation. Especially after the Financial Crisis.

109 Figure 10.11

Jacobsen & Naug Model Vs. House price index (percent change over four quarter period) 1990-2016 25% 20% 15% 10% 5% 0% -5% -10% -15%

J&N (% Change) House price index

Source: Naug, 2017; NCB, 2017; Own calculations

One of the larger critiques of the model is that during the authors estimation period of 1990-2004 Norway was in a stable financial period with minor changes in the economy. An economic model should know how to deal with an economic crisis when it appears. Although the authors added an expectation factor in the model, it has not been able to adjust for a crisis. We can see the results of this in figure 10.2, which shows our estimation period. The model seems to deviate from the index to a larger extent in the time after the Financial Crisis.

Another weakness of the model is that is does not catch the price development in urban areas. Although this is difficult as it looks at a national price development, it is important to point out that in general the price development has been much stronger in urban areas compared to rural areas. This could lead to bubbles in urban areas, which will not be caught by the model.

Earlier we said we would discuss closer the effect of the interest rate, both for t and t-1. We find it interesting that the interest rate effects housing prices to such a large extent at period t compared to t-1. The authors state that an interest rate change seems to effect housing prices more in the short term, rather than the long term. We would think that such a macroeconomic change would take more than one period

110 to settle and thus effect housing prices more in t-1. However, these are the results and our task is to interpret them.

Also, it is unclear to us why the authors chose to use the unemployment rate at period t in the square brackets, when all other coefficients were used in t-1. We see no reason to believe that a change in the unemployment rate will affect housing prices at a quicker rate than a change in interest rate will.

Lastly, we thought it might have been beneficial for the model to include more error adjustment terms, such as the existing square bracket, only with coefficients with more lagged terms. As it stands now the lag only explains changes in the last period. However, for quarterly data it could be beneficial to include more lags. Especially since this is a house price model, where some of the variables have slow long-run effects.

10.6. Conclusion The model that Jacobsen and Naug introduced in 2004 explained to a large extent the price increase seen during their period. Also, it was used to predict future movements. However, when we look at a longer time-period that includes many different economic situations, the model loses some of its explanatory power. The most important factors during the original time period still explain much of the increase seen in our period, but the coefficients of the variables have changed. As the values of the coefficients vary when we changed the estimation period, the model seems to be volatile to changes in the economy. Analyzing the model’s predictions with the actual movement in housing prices also shows that the model has trouble with explaining recent price developments. This indicates that the model is not able to precisely predict future developments in housing prices. From this, we can conclude that the model has not adapted well to the current economic situation and should be used with caution.

111 11. Case and Schiller’s Seven Criteria for a Housing Bubble As there is a lot of psychology involved in the housing market, and thus how the prices evolve, we will try to analyze some of the psychological factors that Case and Schiller name as the most important signs of a bubble. These are harder to quantify and measure accurately, but by analyzing certain indicators in the market we will try to get an overview of the market psychology.

11.1. See housing as an investment In a survey done by the Norwegian Fund and Asset Management Association (NFAMA) in May 2016, 39% of Norwegians saw their own dwelling as their most profitable investment in the next 12 months. Also, 11% ranked real estate, other than their primary dwelling, as their best investment. Real estate investments were therefore ranked as the most profitable by the majority of the responders. When evaluating the riskiness of their investments, housing was ranked as the least risky investment, even safer than bank deposits (Dagens Næringsliv, 2016b). From traditional finance theory, high risk yields high returns and not the other way around. Thus, this survey highlights the irrationality people exhibit when they see their dwelling as an investment. According to the director of NFAMA, Bernt S. Zakkarissen, this has been an increasing trend over the last few years. Zakkarissen said that, "people see housing as a magic investment that combines the highest reward with the lowest risk" (Dagens Næringsliv, 2016b). Holbergfondene states that 54% of the total household economy is placed in real estate, while only 16% is placed in stocks, funds or other financial savings. There has never been a period with a larger growth in house prices than the last 25 years, yet the return on investment has been higher for stocks and equally good for bonds (Holbergfondene, 2017). Secondary housing is an even stronger indicator, though it is difficult to measure whether the motivation is capital gains or consumer needs. In 2015 the total number of so-called secondary houses was 288,000, with a total value of 304 billion NOK (VG, 2015). From 2010 to 2013 there was a 22% increase in Oslo, giving a total of 52,300 secondary houses in Oslo in 2013. For those buying a second dwelling, the number one reason is the hope of making a good investment (NRK, 2015). To make it harder to buy a second dwelling in Oslo, the government has set stricter rules for the financing of these objects. We still must wait and see if the effects of these new rules will be as intended by the Minister of Finance.

112 11.2. Widespread agreement of an increase in prices One way to quantify and test for an agreement of an increase in housing prices is to look at the amount of dwellings sold. Actual number of sales has been quite stable for the last 10 years, implying that there at least has not been a fear for a decrease in prices. An increase in sales could hint at a decreasing price, allowing more people into the market. Except for the years during the Financial Crisis, the number of sales has been around 80,000, as seen in Appendix 1.1. Thus, number of sales show no sign of people fearing declining prices. In an attempt to control the extreme price increase in Oslo, the government introduced temporary regulations on banks’ lending-policies. This seems to indicate that the government agrees that prices, especially in Oslo, will continue to increase. An increase in prices could also be a self- fulfilling prophecy, where the expectation of increased prices makes the prices go up. This effect can of course work both ways; if the market agrees that prices will go down it may cause an actual drop in prices.

11.3. Exaggerated expectations, excitement and word of mouth By searching for "the Norwegian housing market" (boligmarkedet i Norge) on Google we get approximately 315 000 matches, with more than 25 600 news-articles (searched 05.04.2017). Each day a new article trying to predict the development of the Norwegian housing market is published. Limiting our search to news articles only from the past year still yields 5050 matches on Google, showing it is a hot topic. As the media tends to write about the topics that attracts more readers, it is likely that housing and housing prices is something that is important and interesting for the average Norwegian. An example of this is found in a gathering of front pages of two of the most important financial newspapers in Norway. This is found Appendix 1.2.

11.4. Sense of urgency in buying a home Nobody wants to miss out on the opportunity to "gain" from the increase in the housing market, and therefore people feel a sense of urgency in buying a home. This has forced a lot of first time home buyers to get help from their parents to be able to enter the housing market and to buy their first home. Parents and grandparents sacrifice their own investments to help their kids, while some also see it as a way of investing their own money in a “safe bet” (Dagens Næringsliv, 2015b).

113

In Trondheim, Boligbyggelaget Tobb have created so called "renting-before-owning" apartments, for first time buyers to be able to buy without fulfilling the requirement of 15% equity. After renting for 3 years, they can buy the apartment without any equity. This project got more than 400 interested, and all were between the age of 25 and 35 (Hegnar, 2017).This is a fairly good indicator for the stress a lot of young homebuyers are feeling, and the urgency to buy a home.

11.5. Simple or simplistic theories The amount of forced sales can give an indication to how people sometime use simplified assumptions about the economic situation and the housing market. Most people have to take on a lot of debt in order to buy a dwelling. If they fail to service their mortgage, the banks will put their dwellings up for a forced sale. They might have taken on too much risk, or not been realistic about their own economic situation and the duration of their investment. The number of forced sales seems to have been stable since the beginning of the 2000's until the Financial Crisis. Now it seems to have stabilized at a slightly higher level than what it was prior to the crisis, which therefore hints to more simplistic and unrealistic theories and expectations among the population. A figure of this development is found in Appendix 1.3.

11.6. The occurrences of sales above asking prices For the past 14 years, sales above the asking price has been a common occurrence when buying a house. Except for a significant drop in 2009 and a small in mid-2003, the yearly average deviation of price to asking price has been above zero every year since 2003 (Eiendom Norge, 2017a).

114 Figure 11.1

Source: Eiendom Norge, Finn, Eiendomsverdi AS

There has been a problem with so-called underpricing, where real estate agents underprice to get more interested parties to come to dwelling viewings and to join the bidding rounds. This can cause the price to potentially increase a lot. The Norwegian Consumer Council (Forbrukerrådet) has tried to control this behavior, to get the extreme bidding wars under control. The occurrence of sales above asking prices does not necessarily have to indicate a bubble. It might just be a product of excess demand or that the real estate agents deliberately underprice. Still, these are both signs of an unhealthy market, and these could in turn create bubble tendencies.

11.7. Perception of risk As stated in the discussion earlier, people see housing as their most profitable and safest investment. Thus, people do not fully understand the risks of buying a dwelling. Some forget about the fact that housing prices can go down, and that they have done so several times when you look at the history of the housing market both in Norway and abroad. Others forget that they must be able to service their mortgage. With the historically low key rate that we have now, there is not much room for it to continue go down, but it has a lot of room to increase. An increase of the key rate might render people unable to

115 service their mortgage. So far, NCB’s latest projections for the immediate future is to keep the key rate at roughly the same level (NCB, 2017). This projection was shown in figure 8.3 in section 8.

11.8. Conclusion It seems that most of the seven characteristics are pointing towards a bubble in the Norwegian housing market. A reason for this is that they are closely linked together. A widespread agreement of an increase in housing prices will give people an incentive to quickly enter the market to benefit from the expected price increase. And with little perceived risk, people get an even stronger sense of urgency as they see no real downside with entering the housing market. This will again lead to more sales above asking prices, as the pressure on the houses for sale increases. Together, this creates a stable and high number of sales each year. One problem with applying this logic to the Norwegian housing market, is that this has been the situation for most of the indicators for more than 10 years. Therefore, it is hard to tell whether the indicators created for the American market can be easily applied to the Norwegian market. Finding data to analyze and test the indicators is also difficult, and this weakens the power of the analyses done here. The strongest bubble sign we see after analyzing these indicators seems to be the fact that the government has implemented rules to try to control the price increase and regulate the housing market, and the high level of media coverage this topic has received. Lately, the price increase has shown some signs of cooling down. This could very well be the first step towards a healthier growth in prices, and not necessarily a sign that a housing bubble is going to burst.

116 12. What do the Experts and Professionals Say? Throughout our process of writing this master thesis we have without doubt made up our own conclusions about whether the Norwegian housing market is currently in a housing bubble or not. This will be written about in further detail in our final conclusion in the next section. Before that, we thought it might be interesting to investigate what various experts and professionals who work with this topic every day think.

Common for most of the experts and professionals we have found quoted in articles and that we personally have talked to is that there is almost no one who actually believes we have a housing bubble in Norway. Rather than a bubble crash is that we will experience a “soft landing”. The governor of the Norwegian Central Bank, Øystein Olsen, is convinced of this because of the precautions the Norwegian government has made (E24, 2017).

With regards to the fundamental factors we have analyzed throughout this thesis we thought it would be interesting to see which factors other professionals and experts thought are most explanatory to the housing price increase. The highly renowned real estate broker, Grethe Meier, listed these factors, “A low interest rate has been the most influential. Also, it has been easy to get financing from the bank, and with a low unemployment rate along with other macro-numbers which have in general moved in the right direction, this has pushed the housing market up. Also, the housing market is affected by a lot of psychology, and this is only amplified when house prices increase as more people want to enter to reap the benefits, pushing prices even higher” (Meier, 2017). The same factors are backed up by CEO of Eiendom Norge, Christian Dreyer. He also includes the fact that there has been built far too few dwellings in recent time (Dreyer, 2017).

The last part of what Grethe Meier said caught our eye, because this could by definition be viewed as bubble tendencies. However, CEO of Eiendom Norge, Christian Dreyer, had this to say about the bubble discussion, “The last year we have seen an increase of investors who buy only for a re-sale or to rent out. Many of these buyers buy based on an expectation of further price growth; for many the very definition of a bubble. However, the discussion around if there is a bubble or not is often disturbing. The

117 most important part is if we can sustain the overall debt burden, something I believe we can” (Dreyer, 2017).

Lastly, we asked them about the market and whether or not they thought we currently were facing a bubble. This is what they had to say, “There is no housing bubble now. In Oslo where prices have gone up as much as they have there is still too few new builds such that it is not possible to experience a large drop. What we are seeing now is that all these new regulations are effecting the market negatively. The market is held up and fewer dwellings are sold, especially in Oslo because many want to wait and see the effect of the new mortgage-regulations. There will without doubt be a flatter growth-rate, but I don’t think there will be a crash” (Meier, 2017). Chief analyst in Sparebank 1 in Norway has the following to say about the situation, “Housing prices seem to be a little overpriced and is currently at a very high level. However, throughout our many analyses we cannot conclude that there is a bubble in the market. I see no danger signs of that we will experience a housing crash in Norway” (Dagens Næringsliv, 2017).

After talking to several experts and professionals we feel like we have only further backed up our own conclusions for our master thesis. Furthermore, having these experts give us their latest thoughts provides a more solid foundation for our analysis, since they all work with the topic each day and know the market extremely well. Therefore, we will have their opinions in mind when we will state our final conclusion in the next section.

118 13. Conclusion By looking at the historical development of the Norwegian housing market and comparing it to the development seen in the last 25 years, it is natural to assume that prices will drop in the near future. That is, all periods with excessive growth have been followed by a significant drop in prices. The intention of this master’s thesis has been to investigate if there currently is a bubble in the Norwegian housing market.

By conducting a comparative empirical analysis, we were able to better understand the Norwegian market in relation to its fellow Scandinavian countries. The HP filter showed that prices in Norway are not above the long-term trend. However, both for Sweden and Denmark prices are above their long-term trend. Hence, showing bubble tendencies. By utilizing the P/R ratio housing prices in Norway seem to be high compared to rent prices. This is observed by the real P/R ratio, which now is close to 20. Also, this is seen as the real P/R ratio lies above the fundamental P/R ratio. Compared to Sweden, where the gap between the real and fundamental P/R ratio is large, the deviation is relatively small for Norway. From this, the Norwegian housing market seems to be less vulnerable than the Swedish market. Tobin’s Q supports the findings of high housing prices, as its current ratio of 1.08 is above its equilibrium. However, historically this ratio is seen to be highly volatile.

Analyzing the fundamental factors on both the supply and demand side of the housing market, we established more of an in depth understanding of how various macroeconomic factors affect the housing market. A record low key rate, has for many years supported a price increase in the housing market. A low key rate directly improves the purchasing power of each household, thus an increased price level is supported. Equally important is the fact that NCB predicts a low key rate for years to come, further backing the price increase. For many years now we have experienced a shortage in the housing stock, with the amount of new builds being lower than the demand each year. This is emphasized by the fact that 2016 was the first year since the beginning of the 1980’s that more than 35,000 new builds were initiated. Further driving the excess demand is the population growth Norway has experienced. The last 20 years, the population has increased by almost one million, where net immigration has played a substantial role. This is a situation that favors an increase in housing prices now and for years to come. Compared to the Scandinavian countries, Norway has maintained a low and stable unemployment rate

119 the last 25 years. This is clearly a factor that affects the price level for housing in a positive way. Lastly, all these positive growth rate factors contribute to raising the expectations of increased prices in the future. The expectations factor is in itself equally important for further price development as the other fundamental factors.

Our re-estimate of the Jacobsen & Naug model showed the explanatory power had dropped to 74%, compared to the original 88%. The coefficients still affect housing prices in the same direction, although to different extents. Given this, and as housing prices have continued to increase since they created the model, we can conclude that there are more factors now that positively affect housing prices which are not included in the model.

Applying Case & Schiller’s seven criteria for a housing bubble showed that many, if not all, characteristics pointed towards a bubble. However, measuring market psychology accurately is extremely difficult and should therefore only be used as guidance. Receiving direct information from various experts and professionals in the market gave us a better indication of how the current situation actually is. They all reached the same conclusion, that there currently is no bubble in the housing market.

Through empirical and descriptive analyses throughout this master thesis, we have gained a deeper understanding of the housing market. From this, we feel confident enough to state a conclusion. Although the housing price level is high, and the growth has been high for a long time, we see no reason to believe that we are currently experiencing a housing bubble that will burst, resulting in a significant price drop. We believe that the fundamental factors will support a further price increase, though at a much lower rate than before.

120 14. References ABC Nyheter (2017). Svensk boligmarked nærmest kollaps i Europa. Retrieved 17.03.17

Backus, D., K., & Kehoe, P., J. (1992). International Evidence of the Historical Properties of Business Cycles. The American Economic Review, Vol. 82 No. 4, 864-888

Bloomberg (2016). Denmark Faces “Out of Control” Housing Market. Retrieved 15.02.17

Boug, P. & Dyvi, Y. (2008). En makroøkonomisk modell for norsk økonomi. Sosiale og økonomiske studier 111, updated 2008, p. 191-199

Brooks, Chris (2008). Introductory Econometrics for Finance. 2nd edition, Cambridge, United Kingdom: University printing House

Case, K. E. & Shiller, R. J (2004). Is there a Bubble in the Housing Market? 34 (2004-2): 299-362

Cembrit (2015). NYHED: Byggeomkostninger spurter fra huspriser. Retrieved 04.03.17

Corgel, J. B. (1997). Determinants of hotel property prices. Cornell University, School of Hotel Administration

Correia, I., H., Neves, J., L. & Rebelo, S., T. (1992). Business Cycles from 1850-1950: New Facts about Old Data. European Economic Review, 36:2/3, 459-467

121 Creswell, John W. (2014). Research Design: Qualitative, Quantitative and Mixed Approaches. Part 1.

Cuthbertson, K. & Nitzche, D. (2004). Quantitative Financial Economics. 2nd edition, chapter 2

Dagens Næringsliv (2014). Tilbud og etterspørsel. Retrieved 03.03.17

Dagens Næringsliv (2015a). Ap mener Høyre blåser opp boligtall. Retrieved 02.04.17

Dagens Næringsliv (2015b). Foreldre dropper egne drømmer for å hjelpe barna. Retrieved 23.03.17

Dagens Næringsliv (2016a). Det går ikke an at dette fortsetter de neste årene, det er helt umulig. Retrieved 02.02.17

Dagens Næringsliv (2016b). -Sterkere tro på at bolig er en magisk investering. Retrieved 23.03.17

Dagens Næringsliv (2017). Ny boligprisrapport: Boliger er overpriset, men Norge har ingen boligboble. Retrieved 18.01.17

DR Nyheter (2016). Nationalbanken ser tegn på ny mulig boligboble I København. Retrieved 03.04.17

122

Dreyer, Christian V. (2017). Email correspondence, see Appendix 1.5.

DST (2016). Prisindeks for Ejendomssalg. Retrieved 28.03.17.

DST (2017). Byggeomkostningsindeks for boliger. Retrieved 28.03.17

Eiendom Norge (2013). Når alle skal bo og alle vil eie. Retrieved 18.03.17

Eiendom Norge (2016). Boligprisstatistikk. Desember 2016. Eiendom Norge, 1-12

Eiendom Norge (2017a). Boligprisstatistikk. Mars 2017. Eiendom Norge

Eiendom Norge (2017b). Boligprisstatistikk. Årsrapport 2016. Eiendom Norge

Ensliges Landsforbund (2014). Hvor bor de enslige?. Retrieved 04.04.17

European Central Bank (2005). A Trend-Cycle(-Season) Filter, Working Paper Series no. 499. 1-10

European Central Bank (2013). Economic and Monetary Developments. Monthly Bulletin, May 2013, 65-67

123 E24 (2016). Så mange Oslo-boliger må bygges for å løse boligmangelen. Retrieved 27.03.17

E24 (2017). Boligprisene begynner og bite. Retrieved 27.03.17

Foote, C. (2010). Investment and the Housing Market. Lecture from Harvard

FSAN (2011a). Retningslinjer for forsvarlig utlånspraksis for lån til boligformål, The Financial Supervisory Authority of Norway, 29/2011, 1-5

FSAN (2011b). Boliglånsundersøkelsen. The Financial Supervisory Authority of Norway, 2011, 1-14

FSAN, (2012). Boliglånsundersøkelsen. The Financial Supervisory Authority of Norway, 2012, 1-14

FSAN (2015). Finanstilsynet foreslår å forskriftsfeste krav til utlånspraksis for boliglån. Retrieved 20.03.17

FSAN (2016a). Boliglånsundersøkelsen. The Financial Supervisory Authority of Norway, 2015, 1-12

FSAN (2016b). Finanstilsynet foreslår innstramminger i boliglånsforskriften, Retrieved 14.02.17

Gordon, M., J. & Shapiro, E. (1956). Capital equipment analysis: The required rate of profit. Management Science, Vol. 3 No. 1, 102-110

124 Gram, T. (2011). Når staten tar kontroll. Norwegian Central Bank, Staff Memos, No. 18, 1-118

Grytten, O. H. (2002). Norwegian Policy Response to the International Economic Disintegration during the Inter-War Years. Discussion Paper

Grytten, O. H. (2009). Boligboble? - Empiriske indikatorer i historisk perspektiv. Magma.

Hamilton, J., D. (2017). Why You Should Never Use the Hodrick-Prescott Filter. Department of Economics UC San Diego, 1-48

Hayashi, F. (1982). Tobin’s Marginal Q and Average Q: A Neoclassical Interpretation. Econometrica, 213-224

Hegnar (2017). Stor interesse for «leie-før-eie» leiligheter. Retrieved 23.03.17

Hendry, D. F. (1984). Econometric Modelling of House Prices in the United Kingdom, in Hendry, D. F. &Wallis, K.F. (eds). Econometrics and Qualitative Economics, Oxford: Basil Blackwell, 211-252

Hodrick, R., J. & Prescott, E., C. (1981). Postwar U.S. Business Cycles: An Empirical Investigation. Center for Mathematical Studies in Economics and Management Science, Discussion Papers 451

Hodrick, R., J. & Prescott, E., C. (1997). Postwar U.S. Business Cycles: An Empirical Investigation. Journal of Money, Credit, and Banking, Vol. 29 No. 1, 1-16

Holbergfondene (01.27.17). Holberggrafene, Holbergfondene

Husbanken (1985-2015). Historisk statistikk. Retrieved 15.02.17.

125 Jacobsen, D. H. & Naug, B. E. (2004). What drives house prices?. Norwegian Central Ban, Economic Bulletin 05 Q1, 29-41

Kartverket (2017). Dokumentavgift. Retrieved 02.04.17

Kim, H. (2004). Hodrick-Prescott Filter. Auburn University, 1-1

Larsen, E., R. (2005). Boligprisenes utvikling. Økonomiske analyser, No. 5, 26-33

Lerbs, O. W. (2012). House prices, housing development costs and the supply of new single-family housing in German countries and cities. Mü nster: Center of Applied Economic Research Mü nster, University of Mü nster, No. 57 Working Paper (ed.), 7

Lovdata (2017). Lov om planlegging og byggesaksbehandling, Retrieved 28.03.17

Meier, Grethe W. (2017). Email correspondence, see Appendix 1.4.

Miller, M., H., & Modigliani, F. (1961). Dividend policy, growth, and the valuation of shares. The Journal of Business, Vol. 34 No. 4, 411-433

Mohr, M. (2005). A Trend-Cycle(-Season) Filter. European Central Bank, Working Paper Series, No. 499, 1-42

Naug, B. E. (2017). Received dataset via email, 28.02.2017

NCB (2003). Finansiell stabilitet. The Norwegian Central Bank Rapport Series, No. 2, 1-45

126 NCB (2013). Key indicators for a countercyclical capital buffer in Norway – Trends and uncertainty, The Norwegian Central Bank, Staff Memo No. 13, 1-41

NCB (2017). House price Indices. Retrieved 20.03.17

NEF (2013). Historiske priser 1985-2013. Retrieved 13.02.17

NOU (2002). Boligmarkedene og boligpolitikken. Norges offentlige utredninger, 2002:2, 1-411

NRK (2013). Byggekostnadene har eksplodert. Retrieved 15.03.17

NRK (2015). «Hytter» i storbyene presser opp boligmarkedet. Retrieved 19.03.17

Organization for Economic Co-operation and Development (2006). Recent house price developments: The role of fundamentals. Economics department working papers, No. 475, 1-54

Pearson, K. (1895). Notes on regression and inheritance in the case of two parents. Proceedings of the Royal Society of London, Vol. 58, 240-242

Poterba, J., M. (1984). Tax Subsidies to Owner-Occupied Housing: An Asset-Market Approach. The Quarterly Journal of Economics, Vol. 99 No. 4, 729-752

Ravn, M., O., & Uhlig, H. (2002). Notes: On adjusting the Hodrick-Prescott filter for the frequency of observations. The Review of Economics and Statistics, Vol. 84 No. 2, 371-380

127 Regjeringen (2016). Fastsetter ny boliglånsforskrift. Retrieved 15.01.17

Romer, C., D. (1991). Changes in Business Cycles: Evidence and Explanations. Journal of Economic Perspectives, Vol. 13 No. 2, 23-44

Rødseth, A. (1987). Bustadsmarknaden – utviklingstrekk og verkemåte. Sosialøkonomen, Nr. 11, 8-16

SCB (2015). Rents for dwellings 2014. Statistiska Centralbyrån, 1-41

SCB (2017a). Consumer price index, Rent. Retrieved 24.03.17

SCB (2017b). Fastighetsprisindex. Retrieved 24.03.17

SCB (2017c). Lending rates to households and non-financial corporations. Retrieved 24.03.17

Skatteetaten (2017a). Tax deduction card and assessed value of homes. Retrieved 15.03.17

Skatteetaten (2017b). Tax on the letting of housing and holiday homes. Retrieved 15.03.17

128

Skatteetaten (2017c). Tax value of residential properties. Retrieved 15.03.17

Skatteverket (2016). Tax rates. Retrieved 24.03.17

SSB (2011). Hva driver utviklingen i boligprisene?. Retrieved 29.03.17

SSB (2014). Land use and land cover. Retrieved 29.03.17

SSB (2016a). Boforhold, levekårsundersøkelsen, 2015. Retreived 30.03.17

SSB (2016b). Boforhold, registerbasert, 1. Januar 2015. Retrieved 02.04.17

SSB (2016c). Eiendomsskatt 2016. Retrived 02.04.17

SSB (2016d). Population and land area in urban settlements, 1 January 2016. Retrieved 03.04.17

SSB (2017a). Alminnelig inntekt. Retrieved 20.03.17

129

SSB (2017b). Byggeareal. Retrieved 21.03.17

SSB (2017c). Construction cost index for residential buildings. Retrieved 22.03.17

SSB (2017d). Konsumprisindeksen. Retrieved 23.03.17

SSB (2017e). Leiemarkedsundersøkelsen. Retrieved 24.02.17

SSB (2017f). Nøkkeltall for befolkning. Retrieved from

SSB (2017g). Prisindeks for brukte boliger. Retrieved 13.02.17

SSB (2017h). Renter i banker og kredittforetak. Retrieved 20.03.17

130 =BankUtInnRent&nvl=&PLanguage=0&nyTmpVar=true&CMSSubjectArea=bank-og- finansmarked&KortNavnWeb=renter&StatVariant=&checked=true>

SSB (2017i). Eiendomsomsetning. Retrieved 29.04.17

Stiglitz, J. E. (1990). Symposium on bubbles. The Journal of Economic Perspectives, Vol. 4, No. 2. (Spring, 1990), 13-18.

SvD Naringsliv (2017). Nytt prisrekord – trycket stiger på bomarknaden. Retrieved 15.02.17

Svensk Maklarstatistik (2017). Prisutveckling. Retrieved 27.02.17

Søbye, E. (2000). Kristianiakrakket 1899. Statistikk & historie – Statistiske Analyser, Chapter 8.

TIME (2010). Price-to-Rent Ratio: Where Houses are the Best Buy. Retrieved 25.03.17

Tobin, J. (1969). A General Equilibrium Approach to Monetary Theory. Journal of Money, Credit and Banking Vol. 1, No. 1

Torsvik, R., M. (1999). Bankkrisen. Statistikk mot år 2000: 1990-1991, Statistisk Sentralbyrå

Tufte, E., R. (2006). The Cognitive Style of PowerPoint: Pitching Out Corrupts Within. Cheshire, Connecticut: Graphics Press., 2nd Ed.

131 VG (2015). Nye tall: Nordmenn har bolig nummer to for over 300 milliarder. Retrieved 22.03.17

132 15. Appendix A

Appendix 1.1

Source: Eiendom Norge, 2017b

133 Appendix 1.2

Source: Holbergfondene 2017

Appendix 1.3

Forced Sales 250

200

150

100

50

0

2000K1 2000K3 2001K1 2001K3 2002K1 2002K3 2003K1 2003K3 2004K1 2004K3 2005K1 2005K3 2006K1 2006K3 2007K1 2007K3 2008K1 2008K3 2009K1 2009K3 2010K1 2010K3 2011K1 2011K3 2012K1 2012K3 2013K1 2013K3 2014K1 2014K3 2015K1 2015K3 2016K1 2016K3 2017K1 Source: SSB, 2017i

134 Appendix 1.4

135 Appendix 1.5

136