The German house price puzzle

(1)

.

Amsterdam Business School

Thesis

To obtain the academic degree

Master of Science (M.Sc.) in Business Economics: Real Estate Finance

The German House Price Puzzle

Name: Wolfgang Will

Student-Nr.: 10825711

Thesis Supervisor: Prof. Dr. M.K. Francke Second Reader: Dr. M.I. Dröes

(2)

.

Statement of Originality

This document is written by Student Wolfgang Will who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used creating it. The Faculty of Economics and Business is responsible solely for the supervision of

(3)

The German House Price Puzzle .

Abstract

Over the past decades, international housing prices experienced extensive appreciation and depreciation phases, in particular prior and following the financial crisis in 2008. In stark contrast, real house prices in Germany stagnated over more than 40 years, experiencing significant stability. While available literature acknowledges the existence of a German house price puzzle, literature also acknowledges the evident lack of research on explaining the divergent and subdued behaviour of German house prices. Therefore, this thesis attempts to, first, empirically quantify the link between the macro economy (as a principal determinant of house prices) and German house prices, and second, to identify specifics of the German housing market that ultimately explain the divergence from international housing markets with The Netherlands as the subject of comparison. The findings indicate that German house prices are only weakly related to changes in macroeconomic fundamentals, which is found to be strongly associated with the specifics of the financing and rental market. More explicitly, in particular the observed low level of indebtedness and low rate of homeownership isolate the German housing market from distortions in the macro economy and contribute significantly to the house price stability, explaining the divergence from The Netherlands.

(5)

The German House Price Puzzle ii

List of Abbreviations

ADF Augmented Dickey-Fuller AIC Akaike Information Criterion ARDL Autoregressive-Distributed Lag BIC Bayesian Information Criterion CBS Centraal Bureau voor de Statistiek CPI Consumer Price Index

Destatis Statistische Bundesamt DGZF Deutsche Girozentrale Fixing ECM Error-Correction Mechanism

EU European Union

GDP Gross Domestic Product IRF Impulse-Response Function LTV Loan-to-Value

NHG Nationale Hypotheek Garantie

OECD Organisation for Economic Co-operation & Development OLS Ordinary Least Squares

PfandBG Pfandbriefgesetz

UK United Kingdom

US United States

VAR Vector Auto-Regression

(6)

The German House Price Puzzle iii

Table of Figures and Tables

Figure 1: Real OECD House Price Index ... 2

Figure 2: Error-Correction Terms ... 30

Figure 3: Total Outstanding Residential Loans ... 39

Figure 4: Adjusted Total Outstanding Residential Loan Index ... 39

Figure 5: Total Outstanding Residential Loans per Capita ... 40

Figure 6: Total Outstanding Residential Loans to Disposable Income Ratio ... 41

Figure 7: Homeownership Rates ... 43

Figure 8: Available Dwelling Stock per Capita ... 44

Figure 9: Price-to-Rent Ratio Index ... 45

Figure 10: Long-Run Price Elasticity of New Housing Supply ... 48

Figure 11: Number of Issued Building Permits ... 49

Figure 12: Number of Completed Housing Projects to Dwelling Stock Ratio ... 50

Figure 13: Number of Transactions of New or Second-Hand Dwellings ... 51

Figure 14: Number of Transactions to Dwelling Stock Ratio ... 52

Figure 15: House Price Index for Germany, West Germany, and East Germany ... 53

Figure 16: House Price Index for Germany and 16 German Federal States ... 54

Table 1: Selected Literature on the Role of Macroeconomic Fundamentals ... 10

Table 2: Research Hypotheses ... 17

Table 3: Expected Signs of Coefficients of Selected Explanatory Variables ... 19

Table 4: Macroeconomic Fundamentals Employed in the Statistical Analysis ... 23

Table 5: Cross-Country House-Price Level Correlation Matrix ... 25

Table 6: Descriptive Statistics on Employed Variables in Levels ... 26

Table 7: Descriptive Statistics on Employed Variables in First Differences ... 26

Table 8: Engle-Granger ECM Regression Output I ... 28

Table 9: ADF Unit Root Test on Residuals ... 31

Table 10: Engle-Granger ECM Regression Output II ... 33

Table 11: F-Test on Long- and Short-Run Interaction Terms ... 35

Table 12: Direct Comparison of Mortgage Market Characteristics ... 38

Table 13: Direct Comparison of Regulatory Characteristics ... 46

Table 14: House-Price Correlation Matrix on Sub-National Level ... 53

Table 15: House-Price Correlation Matrix on Federal State Level ... 54

(7)

The German House Price Puzzle 1

1 Introduction

1.1 Origin of the Research Question

The performance of international housing prices was marked by significant appreciation and following depreciation phases over past decades, exposing homeowners, investors, and other stakeholders to extraordinary volatility. On average, real house prices appreciated by 9.2% p.a. in “boom” and depreciated by 7.2% p.a. in “bust” phases1 (Möbert, Peters, & Lechler, 2014). In particular, the most recent (financial) crisis between 2007 and 2009 underlined the divergence between selected housing markets. While the median duration of a boom phase is estimated to be 17 quartiles, the United States (US), Spain, and Ireland experienced respectively a boom in the housing market of 50, 45, and 56 quartiles prior to 2007. During these periods, real house price appreciation of 62% (US), 121% (Spain), and even 251% (Ireland) was recorded, directly followed by a correction of 28%, 35%, and 50%, respectively (Möbert et al., 2014). In stark contrast, no noteworthy boom or bust phases were recorded in Germany for more than 40 years. On average, real house prices increased only marginally by 2.6% p.a. in periods of growth and decreased by 2.3% p.a. in periods of decline (Möbert et al., 2014). Only during the energy crisis in the late 1970s and after the German reunification in the mid 1990s, the German housing market experienced noteworthy dynamics. In these periods, real German house prices increased by up to 12.0%, relative to prices in 1970. However, the appreciation in housing prices in the 1990s was followed by a deflation until 2008, decreasing real house prices to 85.0% of their value in 1970 (Figure 1).

The divergence in house prices becomes particularly puzzling, when directly comparing the German with the Dutch housing market. In contrast to Germany, Dutch house prices experienced two significant boom and bust phases over the same period between 1970 and 2013: First, real house prices increased by up to 92.3% and declined back to par value within a short period of time around 1978 (relative to prices in 1970); Second, and most importantly, real house prices in The Netherlands experienced an extensive growth from 1985 to 2008 of up to 209.7% (relative to prices in 1970), followed by a significant correction in response to the financial crisis

(8)

The German House Price Puzzle 2 (Figure 1). Similarly, the annual changes in Dutch house prices are found to be significantly more volatile, when comparing to changes in German house prices. The instability in prices is denoted by a high standard deviation of 8.5% between 1970 and 2013, in contrast to 2.6% in Germany (Appendix A). The aforementioned findings indicate distinctly different house price dynamics when comparing both countries.

However, according to Micheli, Rouwendal, and Dekkers (2014), Dutch and German house prices should co-move to a significantly greater extent due to the closely related economies as well as the joint coordination of monetary and macro-economic policies prior to as well as following the establishment of the common eurozone. However, the authors find that differences in house prices persist even in adjacent regions near the joint border. Furthermore, Vansteenkiste and Hiebert (2011) argue that co-movement in international housing cycles could be expected for three reasons: co-movement in market fundamentals affecting house prices such as the gross domestic product (GDP), interest rates, or other; common financial market innovations and enhanced financial integration; and convergence of housing-specific factors such as associated risk premia. As expected, the authors find strong positive cross-country linkages within the euro area, however, with the exception of Germany.

Figure 1: Real OECD House Price Index

Notes: The figure illustrates selected OECD real house price indices from 1970 to 2013 comprising Germany (DE), The Netherlands (NL), United Kingdom (UK), United States (US), members of the Organisation for Economic Co-operation and Development (OECD), and countries within the euro area (EURO). The base year of the presented indices is 1970 (=100).

(9)

The German House Price Puzzle 3 Consequently, the question arises on to what extent do market fundamentals determine the particular path of German house prices and what characteristics of the German housing market explain the divergence in prices from other housing markets. A clear understanding of what macroeconomic factors are accountable for the dynamics in German house prices and what housing market specifics yield to the extraordinary house price stability relative to international housing markets is of particular interest for households that are planning to become homeowners or already have achieved homeownership, as well as other stakeholders such as investors and the government. The findings from resolving this German house price puzzle are particularly important against the background of the potential threat arising from continuously decreasing mortgage interest rates. Such an environment could lead to a speculation-driven increase in credit volume and increase in mortgage defaults in case of adverse dynamics in interest rates or other events such as a global economic recession in response to the current euro crisis.

While extensive literature exists on the link between macroeconomic fundamentals and house prices for a variety of countries including Germany, the empirical results on German house prices are oftentimes insignificant or deviate from the results on other housing markets (e.g., Hiebert & Sydow, 2011; & Vansteenkiste & Hiebert, 2011). Hence, Demary (2010) and other researchers advise more detailed elaboration on the diverse empirical responses for Germany relative to other countries. Furthermore, contemporary research emphasises the existence of a German house price puzzle related to the divergent price stability as well as the lack of attention from international journals (e.g., Voigtländer, 2014).

Based on findings presented in literature, theories exist that attempt to explain this German house price puzzle. On the one hand, selected market fundamentals are found to be significantly linked with German house prices, but developed only moderately over time. Hence, house prices developed only moderately likewise (e.g., Kholodilin, Menz, & Siliverstovs, 2008). On the other hand, housing market specifics, such as the persistently low homeownership rate, the conservative mortgage market (e.g., low loan-to-value (LTV) ratios, large share of long-term fixed interest rates, or lack of exotic interest rate terms such as interest-only mortgages), or regulatory

(10)

The German House Price Puzzle 4 framework (e.g., interest rates not tax-deductible), are found to be principal factors in explaining the German house price stability and divergence from other housing markets (e.g., Kofner, 2014; & Voigtländer, 2014). Correspondingly, the aim of this thesis is twofold: first, to assess the role of principal macroeconomic fundamentals in explaining the observed dynamics in German house prices; second, to assess the specifics of the German housing market in explaining the divergence in house prices from other housing markets. Due to the strong historic and contemporary ties on the one hand and significant differences in house price dynamics on the other, the subject of comparison is The Netherlands. The thesis contributes to the existing literature by conducting research on a field of study (i.e., the German house price puzzle) that is not adequately addressed by international journals and lacks a comprehensive explanation (e.g., Voigtländer, 2014). Throughout the course of the research, the thesis combines a statistical analysis of the long- and short-term relationship between macro economy and German house prices with a qualitative analysis of characteristics of the German housing market across multiple dimensions (i.e., financing market, rental market, etc.). By contrasting the findings with findings on the Dutch housing market, the potential factors explaining the German house price puzzle are evaluated and related to findings in literature.

1.2 Research Outline

The remainder of this thesis is structured as follows. Section 2 provides a review of existing related literature. Section 3 defines the employed methodology within the scope of the thesis, with a clear focus on the previously determined research questions. Section 4 presents the empirical data, the construction of key variables, and selected descriptive statistics. Section 5 elaborates on the empirical results of the conducted analysis, following the methodology defined in Section 3. Finally, Section 6 concludes with a discussion of potential limitations of the analysis, the conclusion, and propositions for future research. Further details of the analysis are contained in the appendix.

(11)

2 Literature Review

The following literature review provides a concise summary of relevant research findings with a focus on the German housing market. The review is divided in two parts: First, it presents findings on the relationship between macroeconomic fundamentals and observed housing price dynamics; Second, it presents findings on the role of particular housing market characteristics in explaining the observed divergence in house prices from other housing markets.

2.1 The Role of Macroeconomic Fundamentals in Germany

When assessing the importance of the macro economy with respect to the housing market, fundamentals should include main determinants of demand for and supply of housing. In a stylised environment, the observed dynamics in housing prices are generally the result of endogenous mechanisms (e.g., time lags2) in interaction with exogenous shocks (e.g., shifts in demand), where the latter can have a cyclical or trending nature (Rottke, 2012). More explicitly, Westerheide and Dick (2010) from the Centre for European Economic Research (ZEW) define the following determinants of demand and supply for residential properties: Main determinants of demand are demography (i.e., population, households, & age distribution), income growth, financing conditions, and governmental subsidies; Main determinants of supply are the availability and cost of land, construction costs, investments in existing dwelling stock, and outgoing dwelling stock.

In theory, housing markets are expected to co-move particularly within the euro area due to the common monetary union, as well as increasing degree of integration in trade, financial markets, and economic conditions (Vansteenkiste & Hiebert, 2011). In earlier research, Kasparova and White (2001) investigate the degree of similarities in housing market long- and short-term responses to changes in demand and supply factors across four European countries (including Germany & The Netherlands). The authors employ vector auto-regression (VAR) Granger causality tests to determine the relationship between house prices and GDP as well as the Engle-Granger two-step error-correction mechanism (ECM) with real house prices as

(12)

The German House Price Puzzle 6 the dependent variable and real GDP, real mortgage interest rate, and housing starts as the independent variables on annual data from 1970 to 1998. While concluding that only in Germany GDP does not Granger cause house prices, the authors find that German house prices have a significant long-term relationship with GDP (positive 0.31), mortgage interest rates (negative -0.03), and housing starts (negative -0.16). Further, the error-correction term is estimated to be negative -0.70, which means a rapid correction of 70% within one period (i.e., one year) in case German housing prices and the underlying macroeconomic fundamentals are not in their long-run state of equilibrium. In contrast, GDP is not a significant explanatory of Dutch housing prices and interest rates have an unexpectedly positive and significant coefficient of 0.13 (Kasparova & White, 2001). Further, the speed of adjustment3 within one period is estimated to be 46%. Hence, the Dutch housing market adjusts slower to disequilibria than the German market. While the authors indicate significant long- and short-term differences in housing market behaviour within the euro area, the study covers only observations until 1998 and German data represents West Germany only. Nevertheless, Kasparova and White (2001) propose that it could be of interest to directly compare Germany and The Netherlands, due to the observed divergence in long-run relationships (despite their geographical, economic, and political ties).

Contemporary ECM research results in divergent findings. Koetter and Poghosyan (2010) employ a panel autoregressive-distributed lag (ARDL) ECM on annual data for 78 German economic areas from 1995 to 2004. While GDP per worker (0.63) and population growth (0.16) are significant long-term explanatory variables of house prices, only GDP per worker (-0.30) remains significant in the short term. Further, the speed of adjustment is estimated to be only 10% per year. However, main limitations of the author’s analysis comprise the limited number of explanatory variables, short sample period, and main focus on the impact of house price deviations from equilibrium on German bank stability. Caldera and Johansson (2013) employ the Engle-Granger two-step ECM on quarterly data for 21 country members of the Organisation for Economic Co-operation Development (OECD). For Germany, the contemporary income (0.66), dwelling stock (-1.37), interest rate

(13)

The German House Price Puzzle 7 (0.004), and population (1.08) are found to be highly significant explanatory variables of German house prices and have the expected direction of effect with the exception of the interest rate4 (within the long term). A further essential finding is that the coefficients estimated for Germany are significantly lower than for other OECD countries5_{, which could indicate a subdued responsiveness of German house prices to} exogenous shocks. This observation is also indicated by the small but significant speed of adjustment of 3.5% per quarter (7.9% in The Netherlands). Nevertheless, the author’s main focus is on estimating the long-run price elasticity of new housing supply. Hence, the interpretation of the aforementioned findings is limited. Finally, Agnello and Schuknecht (2011) employ the ECM on annual data for a panel of 18 industrialised countries (including Germany) from 1980 to 2007. While the authors find real GDP growth (+), short-term interest rate (-), and real outstanding credit growth (+) to be highly significant explanatory variables of housing prices and estimate the cross-panel speed of adjustment at 60%, the findings from the panel analysis do not appear to be representative for the German housing market against the background of the aforementioned studies. More importantly, the majority of studies lack an explicit elaboration and interpretation of the divergent findings for Germany.

Tsatsaronis and Zhu (2004) recognise significant differences between individual countries and employ a VAR model across 17 industrialised countries (including Germany) from 1990 to 2003 to assess the house price sensitivity to shocks in inflation, interest rate, and outstanding credit. Focusing on the financing market, the authors find that while inflation is the dominant determinant of changes in house prices, countries with low LTV ratios and a high share of fixed interest rates such as Germany6_{are less sensitive to changes in interest rates than other countries such as} The Netherlands. Goodhart and Hofmann (2008) also employ a VAR model for 17 industrialised countries (including Germany), but on quarterly data from 1970 to

4_{Gürtler and Rehan (2008) provide an explanation for a positive effect on house prices in response to an increase}

in interest rates. According to the authors, an increase in interest rates is generally accompanied by economic growth, hence, a potential increase in demand for housing.

5_{For example in The Netherlands, an increase in income, dwelling stock, interest rate, or population by 1.00%}

results in an change in house prices by 2.52%, 6.61%, -0.005%, and 14.22%, respectively, in the long term, holding else constant. All coefficients are significant at the 5% significance level.

6_{Tsatsaronis and Zhu (2004) assign each of the 17 industrialised countries to one of three groups from most (i.e.,}

group 1) to least conservative mortgage markets (i.e., group 3). Germany is assigned to group one; The Netherlands to group two.

(14)

The German House Price Puzzle 8 2006. The authors find significant multidirectional links between house prices, monetary variables, and the macro economy, where the role of monetary variables (i.e., money & credit) with respect to house prices dynamics is found to be stronger in more recent periods. However, both aforementioned studies present the findings only on an aggregate and not individual-country level. In that regard, Hiebert and Sydow (2011) employ a VAR model for eight euro-area countries (including Germany) from 1978 to 2009 and address findings on an individual level. The authors find that house prices7 across the analysed countries are predominantly explained by variations in disposable income per capita and real interest rates. However, the authors also find that in particular Germany appears as an outlier to the general findings with rather subdued house price dynamics in response to shocks in fundamentals. Similarly, Vansteenkiste and Hiebert (2011) conduct a VAR model on seven euro-area countries (including Germany) from 1971 to 2009, analysing potential spillover effects across European housing markets. Using an impulse-response function (IRF), the authors find relatively low magnitudes of cross-country spillover effects and relatively high market heterogeneity in particular for Germany. German house prices generally observe the lowest response to shocks in European house prices, persistently over multiple periods. More importantly, while a positive shock in domestic long-term interest rates results in an immediate negative effect on house prices in other countries, German house prices respond with a lag of one year, supporting earlier findings of Tsatsaronis and Zhu (2004). Finally, Demary (2010) employs a VAR model on quarterly data for ten OECD countries (including Germany) from 1970 to 2005, finding German house prices to be “sticky” and subdued in response to shocks in the interest rate. In contrast, The Netherlands observes more pronounced effects in response to shocks in fundamentals as well as a direct negative effect on house prices from an increase in interest rates. Nevertheless, Demary (2010) emphasises with reference to Goodhart and Hofmann (2008) the potential shortcomings when comparing empirical results. When conducting a cross-country comparison, varying definitions of the representative property and varying methodologies in collecting the data limit the interpretation of cross-country differences.

(15)

The German House Price Puzzle 9 Alternative research makes a more explicit attempt in identifying the link between macro economy and German house prices as well as explaining the long-term house price stability. For example, Kholodilin et al. (2008) employ a panel regression model on 14 OECD countries from 1974 to 2005 finding that disposable income per capita (+) and level of urbanisation (+) are the most and second most important determinants of house prices, respectively, followed by population growth (+) and interest rates (-). The authors claim that the long-term stagnation in German house prices is the result of German market fundamentals, which developed only moderately over the past decades relative to other OECD countries8. Furthermore, the authors estimate the half-life period of the adjustment speed to be 22 years for Germany (3 years for The Netherlands) resulting in a significantly slower adjustment in house prices in response to changes in underlying fundamentals. According to Möbert et al. (2014), German house prices increased only when the expected income or expected inflation changed significantly. On a regional level, Kajuth, Knetsch, and Pinkwart (2013) and later Bundesbank (2013) find six significant determinants of German housing prices comprising housing stock per capita (-), income per capita (+), unemployment rate (-), population (+), population density (-), and growth expectations (+), where the mortgage rate is found to be not statistically significant (see also Bundesbank, 2014). In result, researchers generally classify Germany as a slow mover (see also Hilbers et al., 2008), when assessing the link between house prices and macro economy. However, Igan et al. (2011) emphasise that the degree in co-movement does not only vary across countries, but also over time, which increases the difficulty of judging estimates and research findings on the German market.

Nevertheless, the available literature and research on the German housing market provides essential findings on the link between housing prices and the macro economy. Table 1 presents a brief summary of the most relevant literature that provides explicit statistical estimates for the German housing market. However, the available literature oftentimes lacks an explicit interpretation of critical observations. First, the role of the financing market, in particular the effect of the mortgage interest rate, is generally not clearly defined and is found to be both positive and negative, as

8_{According to Kholodilin et al. (2008), Germany experienced only low growth in real income, low growth in}

(16)

The German House Price Puzzle 10 well as significant and not significant. Second, data is oftentimes analysed and interpreted on an aggregate level. Third, while the effect of changes in macroeconomic fundamentals on house prices is estimated to be lower relative to other countries (indicated by smaller coefficients), when reported on an individual (i.e., country) level, the corresponding literature oftentimes lacks sufficient elaboration and interpretation of the divergent findings on the German housing market as well as the association with the unique German house price dynamics. Despite several authors recognising the adverse findings for Germany and proposing further related studies, the long- and short-term link between macroeconomic fundamentals and the German house price stability seems yet inadequately clarified. Hence, this thesis contributes to the existing literature by directly assessing the role of principal market fundamentals including the national disposable income per capita, mortgage interest rate, and construction costs in order to explicitly assess their link to the yet puzzling German house price stability.

Table 1: Selected Literature on the Role of Macroeconomic Fundamentals

Notes: The table illustrates a summary of selected literature on the link between macroeconomic fundamentals and German house prices. Estimated coefficients from the respective statistical analysis are stated in parenthesis and only statistically significant coefficients (at the 5% significance level) are presented. The results are for Germany only if not stated otherwise. Long-run variables are employed in levels; short-run variables are employed in first differences. ECT denotes the error-correction term.

(17)

2.2 The Role of German Housing Market Specifics

According to existing literature, while changes in the macro economy denoted by changes in demand- and supply-side fundamentals explain immediate or consequent house price movements to a great extent, particular housing market characteristics explain the divergent magnitude in response to similar exogenous macroeconomic shocks across countries.

The real estate financing market is one of the principal market segments that explain the observed divergence in German house price dynamics. Hiller (2014) determined three characteristics of particularly volatile housing markets comprising low equity (i.e., high LTV) ratios, short-term fixed or floating interest rates, and strong regulations of the housing demand and supply. The author finds the opposite to be the case for the German housing market, which in turn is characterised by high equity ratios and a low interest rate risk9. Typical newly issued German mortgages are characterised by LTV ratios between 60% and 80% (Voigtländer, 2006). Moreover, the aforementioned LTV ratios are based on the long-term achievable value of the property (i.e., lending as opposed to market value), following the prudence concept (Westerheide & Dick, 2010). In result, the assessed lending value can be 10% to 15% lower than the current market value of the property (Westerheide & Rotfuß, 2008), impairing to some extent the comparability of LTV ratios and financing markets across countries (Bundesbank, 2014). More detailed information on specifics of the German in comparison to other mortgage markets can be found for example in Tsatsaronis and Zhu (2004) and Toussaint et al. (2007).

Acknowledging the existence of a German house price puzzle and emphasising the lack of attention from international research, Voigtländer (2014) provides contemporary and highly relevant research on the German house price stability and claims that the German financing market, in conjunction with the well-established rental market, is the principal factor in explaining the observed price stability. According to the author, the aforementioned specifics of the German financing market make the German housing market less prone to macroeconomic

9_{In 2012, 70% of newly issued mortgages included fixed interest rates with a term of over five years, while}

(18)

The German House Price Puzzle 12 shocks, isolating it from financial market distortions, and keeping the risk of default low. Similarly, Kofner (2014) attributes the high quality of the German mortgage market to the tenure structure and persistently low homeownership rate. The low homeownership rate allows financing banks to be more selective in issuing new mortgages, which in turn results in high equity requirements, thorough assessment of the personal creditworthiness, a lack of a subprime mortgage market, low probability of default, and consequently higher house price stability (Kofner, 2014). However, the two authors neither provide statistics on the actual rate of default nor compare Germany to more volatile markets with a similarly low risk of default such as The Netherlands. In that regard, Hiller (2014) emphasises the dominant role of mortgage credit growth as a source of speculative house price inflation, which is confirmed by earlier research conducted by Nguyen (2013), who finds a strong link between mortgage market liberalisation and housing market volatility across 17 OECD countries. According to Kofner (2014) and Bundesbank (2014), such excessive credit growth was not observed in the German housing market and the financing market is regarded rather conservative. Moreover, the primary sources of funding of German financing institutions comprise deposits, bank bonds, and German mortgage Pfandbriefe10, while more exotic funding sources are less common (Kofner, 2014). In conclusion, the existing literature attributes a great extent of the observed German price stability to the specifics of the German financing market.

In addition to the financing market, Voigtländer (2014) emphasises the role of the rental market. In related studies, Boehm and Schlottmann (2014) assess the probability of homeownership in Germany to be unusually low, which is found to be strongly connected to the perception of renting as an attractive alternative to owning (Micheli et al., 2014). Based on interviews among homeowners and tenants in Germany and The Netherlands, conducted for the purpose of assessing the perceived security or insecurity from homeownership, Toussaint et al. (2007) find that particularly young German households prefer rental tenure, due to the availability and high quality of a large rental stock. More importantly, the interviewees emphasise the

10_{The German Pfandbrief is a special form of a covered bond used by financing institutions to refinance issued}

mortgages. In contrast to common covered bonds, the specifics of German Pfandbriefe comprise a statutory regulation within the scope of the German Pfandbrief Act (PfandBG), and a maximum loan value of 60%.

(19)

The German House Price Puzzle 13 affordability of rental stock denoted by low and stable rent levels, uniform statutory tenant protection, and the opportunity to save equity and achieve a secure job before committing to a mortgage that involves a high financial obligation, due to the equity ratio requirements11_{. In contrast, Dutch interviewees generally consider the social} housing sector inaccessible and private rented sector too expensive in The Netherlands. Hence, Dutch households assess the living expenses to be lower when owning relative to renting housing. While the dominant role of the German rental market and large share of private landlords is attributed to Germany’s distinct history resulting in today’s low rate of homeownership (Kofner, 2014; & Voigtländer, 2009), only a limited number of studies directly link the importance of the rental market to the German house price stability. However, the principal claim of contemporary research is that the rate of homeownership is an essential determinant of the German price stability with the argumentation that a low homeownership rate leads to a conservative lending system and not vice versa (Voigtländer, 2014). As stated previously, in a low homeownership environment, financing institutions are able to demand a higher level of creditworthiness from households, maintaining a high level of credit quality and low level of indebtedness (Kofner, 2014). As Kofner (2014) continues, “[…] each additional per cent age point of homeownership rate has to be paid for in the form of worsening average credit quality” (p. 269).

Third, the dominant characteristics of the regulatory framework contributing to the German house price stability include the lack of tax-deductible mortgage interests and low level of homeownership subsidies (e.g., Kofner, 2014; & RICS, 2012). In more detail, Voigtländer (2009) finds the German tax wedge12_on owner-occupied housing to be on average only half of the Dutch, and converging to zero due to the reduction in available subsidies13_{. Moreover, researchers find a variety of} statutory regulations that either encourage long-term investments in housing (e.g., capital gains tax relief after 10-year holding period) or strengthen the rental market (e.g., tax-deductible capital expenditures on rental dwellings, tenant eviction

11_{German households are found to be late home purchasers (Micheli et al., 2014), where the average first-time}

homebuyer is 40 years old (Kofner, 2014). Among other factors, the high equity requirements prohibit in particular young households to achieve homeownership, resulting in the priority of accumulating equity first.

12_{The tax wedge is generally defined as the difference between the nominal mortgage rate and the effective}

after-tax mortgage rate (e.g., Voigtländer, 2009).

(20)

The German House Price Puzzle 14 protection, or legally binding reference rents) (Kofner, 2014). According to Voigtländer (2009), the persistently low homeownership in Germany is the result of such regulations and mainly due to the lack of subsidies, no excessive governmental interventions in rents, a well-established rental housing market, and historic stagnation in house prices. However, while the existing literature relates the regulatory framework to the tenure choice and low homeownership rate, the research lacks the direct association with the German house price stability.

Fourth, German house prices are more stable relative to other countries such as The Netherlands, due to a more responsive housing supply according to selected authors (e.g., Micheli et al., 2014). In order to determine cross-country differences in the responsiveness of housing supply, Caldera and Johansson (2013) estimate the long-run price elasticity of housing supply for 21 OECD countries from the early 1980s to mid-2000s. While the German housing supply increases by approximately 0.43% in response to a 1.00% increase in house prices over the long term, the estimated response of the Dutch housing supply is only 0.19%. According to Voigtländer (2009), the relatively high elasticity is the result of the decentralised planning and issuance of building permits14, which mitigates to a greater extent changes in fundamentals and contributes to the German house price stability. However, the authors do not further elaborate on the link between housing supply and the extraordinary German house price stability, in particular when considering the similarly decentralized planning in The Netherlands or comparing with more volatile, but significantly more responsive housing markets such as the US (2.01).

Fifth, with regards to the investment market, Kofner (2014) finds the German housing transaction market to be relatively less frequented and relates this observation to the specific purchase behaviour of German households. According to the author, Germans are described as “once-in-a-life-time” buyers (see also Pfeiffer & Braun, 2006), whose probability of homeownership as well as the probability of returning to rental tenure or to another owned home once homeownership is achieved is found to

14_{Due to the decentralised planning, German municipalities are competing with each other for citizens and price}

stability. In result, the municipalities are issuing more building permits than in a centralised system, increasing the responsiveness of housing supply and keeping house prices relatively more stable (Voigtländer, 2009).

(21)

The German House Price Puzzle 15 be unusually low (Boehm & Schlottmann, 2014). Additionally, German households are encouraged to be long-term housing investors once homeownership is achieved due to the specifics of the regulatory framework (e.g., capital gains tax waiver), further reducing the number of transactions (e.g., Kofner, 2014). However, the available literature does not specifically relate the specifics of the German investment market to the observed house price stability.

Finally, literature emphasises the limitations and potential of distorted results, when analysing and interpreting data on a highly aggregated level (Kajuth et al., 2013; & Kholodilin, Michelsen, & Ulbricht, 2014). Furthermore, while the German real estate market experienced positive developments with respect to its market transparency, Germany still lacks behind other housing markets, such as the United Kingdom (UK) or US, which experiences a higher level of transparency (Schulte, Rottke, & Pitschke, 2005). According to Rußig and Scharmanski (2004), the evident lack of high-quality long-term data on the German housing market as well as the regional heterogeneity exacerbates the judgement of German house prices. Consequently, Bauer et al. (2013) developed a new hedonic German price index based on data provided by Immobilienscout24 in order to overcome issues arising from traditional highly aggregated price indices. However, while finding significant regional heterogeneity in German house prices, the authors’ analysis is limited to a short sample period, low number of entities, and ask instead of transaction prices. Hence, the results do not provide sufficient evidence for strong regional heterogeneity being a potential factor in explaining the German house price stability.

When assessing the role of specifics of the German housing market in explaining the observed housing price stability based on existing literature, three key observations become evident. First, the specifics of the housing market can be separated into six market segments or dimensions: the financing market, rental market, regulatory framework, housing supply market, investment market, and other, where other denotes issues arising from the available data. Second, while each of the aforementioned dimensions is addressed to greater or lesser extent in existing literature, the availability of research explicitly linking the individual German housing market characteristics to the house price puzzle is limited. Finally, an evident lack of

(22)

The German House Price Puzzle 16 literature exists analysing all six dimensions comprehensively in order to determine valid explanations for the German house price stability. Consequently, this thesis contributes to the existing literature by addressing the aforementioned remarks.

3 Methodology

This thesis develops a two-stage methodology, based on the two previously determined research questions: To what extent do market fundamentals determine the particular path of German house prices and what characteristics of the German housing market explain the divergence in prices from other housing markets? First, the thesis statistically estimates the role of principal macroeconomic fundamentals in explaining German house price dynamics with a focus on both short and long term. In order to evaluate the findings, the same statistical analysis is conducted for the object of comparison, The Netherlands, where both sets of statistical estimates are contrasted against each other. This first stage of analysis aims at quantifying the effect of changes in the macro economy on German house prices and consequently explaining the observed price stability relative to other housing markets. Second, the thesis assesses the role of German housing market characteristics and contrasts the findings to the object of comparison, The Netherlands, by means of a qualitative analysis. This second stage of analysis aims at further explaining in particular the divergence and adverse dynamics in prices relative to other housing markets.

The objective of the proposed methodology is to identify specifics of the German housing market that provide a comprehensive explanation of the German house price puzzle. In result of the existing literature and corresponding research findings, that are predominantly limited to individual parts of the proposed analysis, the following hypotheses are developed and tested to the extent permitted by the available data within Section 5 of this thesis (see Table 2). Hypothesis 1 is tested by statistically analysing the relationship between selected macroeconomic fundamentals and German house prices using quantitative data and employing the well-established Engle-Granger ECM (see Section 3.1). Hypotheses 2 through Hypothesis 7 are tested by qualitatively analysing the specifics of the German housing market, using both qualitative and to some extent quantitative data (see Section 3.2). The hypotheses

(23)

The German House Price Puzzle 17 presented in Table 2 indicate the expected result and are tested for validity in Section 5. However, in particular the test of Hypotheses 2 through 7 is limited to the available data and qualitative evaluation of such. Hence, the results of the analysis are only indicative and need to be treated with care.

Table 2: Research Hypotheses

Notes: The table illustrates the seven hypotheses that are developed in accordance with the literature review and indicate the expected result. The hypotheses are addressed and tested within Section 5 of this thesis.

3.1 Quantitative Analysis of Macroeconomic Fundamentals

For the purpose of estimating the short- and long-term relationship between principal macroeconomic fundamentals and house prices, the thesis employs the well-established two-step ECM developed by Engle and Granger (1987) for each of the two entities: Germany and The Netherlands. Based on the level of importance attributed by existing literature, availability of data, and performance of specific model configurations, two demand-side and one supple-side macroeconomic variables are determined: real national disposable income per capita, real mortgage interest rates, and real construction costs15.

More explicitly, the Engle-Granger cointegration test is a two-step residual-based test. A principal assumption is that all employed variables are integrated of order one, meaning that the variables become stationary by taking their first

15_{For example, additional fundamentals such as the unemployment rate or level of urbanization are statistically}

not significant determinants and decrease the significance of the overall model configuration, while other fundamentals such as the German housing stock or German credit are not available for the entire sample period. Further, a parsimonious model is preferred, in particular due to the limited number of observations.

(24)

The German House Price Puzzle 18 difference. Cointegration exists between non-stationary variables, when a linear combination of the variables exists such that the residuals of this linear combination are stationary themself (i.e., integrated of order zero). Hence, all variables are first tested for their order of integration by means of an augmented Dickey-Fuller (ADF) unit root test. Based on this criterion, the first step of the Engle-Granger ECM includes an ordinary least-squares (OLS) regression model for the purpose of estimating the long-term relationship between the employed variables (Equation 1),

.

ln(hpt) = β0 + β1ln(inct) + β2(intt) + β3ln(cost) + εt (1)

where hpt denotes the real house price level in period t, inct the real national

disposable income per capita, intt the real mortgage interest rate, cost the real

construction costs, and εt the error term. All variables are stated in their natural log

values16. For cointegration between the employed set of variables to exist, the residuals from the first regression model are required to be stationary (i.e., integrated of order zero). Hence, the residuals are tested for stationarity by means of an ADF unit root test. However, the Engel-Granger test statistics do not follow a standard distribution and depend on the number of variables tested as well as the sample size. Hence, critical values specifically developed for the aforementioned test are applied, provided by MacKinnon (2010).

The second step of the Engle-Granger ECM includes the error-correction model, which is estimated using OLS. The one-period lagged residuals from the regression model presented in Equation 1 are employed as an additional independent variable to quantify the short-term speed of adjustment in case of a deviation from the estimated long-run equilibrium. The determination of the employed number of lags within the short-term model presented in Equation 2 is based on the well-established Akaike information criterion (AIC) and Bayesian information criterion (BIC) as well as the annual nature of the underlying data. The number of lags is found to be optimal at one period. Consequently, Equation 2 illustrates the final error-correction model,

.

(25)

The German House Price Puzzle 19 Δln(hpt) = δ0 + δ1Δln(hpt-1) + δ2Δln(inct-1) + δ3Δln(intt-1) + δ4Δln(cost-1) (2) + δ5ECTt-1 + νt .

where the first difference of hp is regressed on the lagged first differences of hp17_{, inc,}

int, and cos, as well as the aforementioned lagged residuals denoted by ECT (i.e., error-correction term). The time series regression models presented in Equation 1 an Equation 2 are conducted separately for both Germany and The Netherlands.

Table 3 illustrates the direction of the expected long-run effects on house prices in response to changes in the independent macroeconomic variables. Based on theory and findings in literature, the two principal demand-side variables disposable income and the mortgage rate are expected to have a positive and negative relationship with house prices, respectively. In other words, if the country’s disposable income or mortgage rate increases, the respective house prices are expected to increase in the former and decrease in the latter case. An increase in the supply-side variable construction costs is expected to have a positive effect on house prices. Finally, house prices are expected to have a significant positive relationship with their own lags, also referred to as inertia or smoothing. More explicitly, contemporary house prices are expected to be partly determined by preceding house price values or changes in values.

Table 3: Expected Signs of Coefficients of Selected Explanatory Variables

Notes: The table illustrates the expected sign of the coefficient (i.e., direction of the effect) on house prices in case of an increase in the corresponding independent macroeconomic variable as well the corresponding rationale.

Hypothesis 1 (Table 2) is ultimately tested by, first, assessing the level of significance of the cointegration between the selected macroeconomic fundamentals

17_{The lagged first difference of hp is included to minimise the autocorrelation in the error term ν. Based on the}

(26)

The German House Price Puzzle 20 and German house prices, and second, by comparing the estimated coefficients on the fundamentals to the estimates for The Netherlands.

3.2 Qualitative Assessment of Housing Market Specifics

In addition to the statistical analysis, a second, predominantly qualitative analysis is conducted in order to explore a comprehensive understanding of the German house price puzzle. Based on findings in literature, the specifics of the German housing market are assessed using a qualitative analysis of six dimensions (i.e., market segments): First, the financing market is analysed with special regards to the development of the mortgage availability, level of indebtedness, share of fixed versus variable mortgage interest-rate terms, and typical LTV ratios. Second, the rental market is assessed with a focus on the rate of homeownership and conditions that favour renting versus owning a residential property, such as changes in the price-to-rent ratio. Third, the regulatory framework is analysed with respect to further conditions favouring renting versus owning, such as specifics of the tax and tenancy laws (e.g., lack of tax-deductible mortgage interest rates or statutory tenant protection). Fourth, the supply of housing is assessed by analysing the supply elasticity and the supply market activity. Fifth, the investment market activity is analysed based on the number of transactions observed in the housing market as well as the housing purchase behaviour. Finally, the thesis analyses other (i.e., alternative) specifics of the German housing market with a particular focus on a potential heterogeneity bias, contained in the national and highly aggregated German house price index. The heterogeneity bias may result due to the significant diversity between West and East Germany or between individual federal states.

Each of the six dimensions is analysed with a clear focus on the German housing market and, subsequently, contrasted with findings on the subject of comparison, The Netherlands. Hypotheses 2 through 7 are ultimately tested by qualitatively evaluating each dimension with respect to its contribution to the German house price puzzle. The principal goal is to identify unique characteristics of Germany that explain the divergence and stability of German house prices relative to the Dutch and other international house price dynamics.

(27)

4 Data

4.1 Employed Data

For the purpose of the first part of the empirical analysis (i.e., the statistical analysis), data on the following seven macroeconomic variables is obtained for the two countries of interest: Germany and The Netherlands.

The OECD provides data on the real house price index (i.e., deflated using the private consumption deflator from the national account statistics; index 1970=100) on a quarterly basis from 1970Q1 to 2013Q4 for both Germany and The Netherlands. According to the data methodology, the German index is based on residential property prices in Germany provided by the Deutsche Bundesbank and the Dutch index is based on the house price index for existing own homes provided by the Kadaster. For the purpose of analysis, the two sets of data are transformed into annual observations.

Furthermore, the OECD provides data on the real net national disposable income (in 2010 euros) as a proxy for the German and Dutch disposable income. Both sets of data are stated on an annual basis from 1970 to 2013. Subsequently, the real disposable income per capita is estimated by dividing the OECD real net national disposable income by the country’s population.

With respect to the German mortgage rate, the Deutsche Girozentrale Fixing (DGZF) Pfandbrief yield curve with a fixed term of ten years18 is commonly used as a proxy for German mortgage interest rates and provided by DekaBank. The data is predominantly stated on a daily basis from 1973 to 2014. For The Netherlands, the mortgage interest rate is characterised by a fixed term of five years19 and provided by TBV Wonen. The data is stated on a monthly basis from 1973M1 to 2013M4. For the purpose of analysis, both German and Dutch mortgage interest rates are transformed into annual data. The real mortgage interest rate is estimated by subtracting the estimated expected inflation20 from the nominal mortgage interest rate.

18_{German mortgages most commonly include a fixed-interest rate with a term of ten years (RICS, 2012).}

19_{In The Netherlands, fixed interest rates with a term of five years used to be most common in the past, but the}

share of shorter interest rate terms increases in popularity (RICS, 2012).

20_{The expected rate of inflation is based on an AR(1) model with a rolling-window estimation. The model is used}

(28)

The German House Price Puzzle 22 As a proxy for German construction costs, Destatis provides the average building costs per m2 for new residential properties on an annual basis from 1962 to 2013. For the Netherlands, CBS provides the construction cost index for new dwellings on a quarterly basis from 1950Q1 to 2014Q3. For the purpose of comparability, the German data series is transformed into a construction cost index and the base year for both index series is set to 1970 (=100). The data is stated in nominal terms and consequently deflated using the construction price deflator provided by Eurostat21.

Eurostat also provides statistics on the level of population for both Germany and The Netherlands. The two sets of data are stated on an annual basis and range from 1960 to 2014. The statistics on the German population include the former German Democratic Republic (i.e., East Germany).

Furthermore, Eurostat provides data on the national consumer price index (CPI) for Germany and The Netherlands on an annual basis from 1960 to 2013. For the purpose of estimating the expected rate of inflation, which in turn is used to estimate the real mortgage interest rate, an autoregressive model of order one (i.e., AR(1)) is defined and applied on rolling window spanning a period of 12 years22. The resulting estimates are used to forecast the historic expected inflation for both Germany and The Netherlands for the required sample period.

Based on the available number of observations for each of the seven macroeconomic variables, the final sample size comprises a total of 40 years of annual observations from 1973 to 201223_{. For illustrative purposes, Table 4 provides a} short summary of the employed macroeconomic fundamentals.

21_{Eurostat provides a construction and building price deflator on an annual basis for both Germany and The}

Netherlands. The deflator is stated as an index (2010=100) for the entire sample period from 1970 to 2013 and based on the euro currency.

22_{The AR(1) configuration resulted in the best and most statistically significant forecast model. Limited to the}

availability of data, a rolling window of 12 years is used. More explicitly, the expected inflation of one period ahead is based on the estimates of the AR(1) model using the 12 preceding periods.

23_{The constraining variable with respect to the sample period is the Dutch mortgage interest rate, which is}

available from 1973M1 to 2013M4. Due to the transformation to annual observations, the final sample period ranges from 1973 to 2012.

(29)

The German House Price Puzzle 23 Table 4: Macroeconomic Fundamentals Employed in the Statistical Analysis

Notes: The table illustrates the seven employed macroeconomic variables, their individual sample period, frequency of observations, and corresponding source for both Germany and The Netherlands.

For the second part (i.e., the qualitative analysis), additional data is obtained for both Germany and The Netherlands. In result, the assessment of the housing market specifics by means of a qualitative analysis across the six defined dimensions is supported by the use of the following quantitative data.

A principal statistical indicator for the financing market is the mortgage availability, which is provided by Hypostat as the outstanding residential loan volume. The data is available for both countries on an annual basis from 1996 to 2013 and stated in nominal terms. Hence, the observations are consequently transformed into real values using the CPI index. Hypostat also provides data on the outstanding residential loans per capita as well as the ratio between outstanding loans and disposable household income. While the former is available for a period from 1996 to 2013, the latter is only available from 2002 to 2013 (both stated annually).

The assessment of the rental market is supported by the use of statistics on the homeownership rate and available dwelling stock. Destatis provides four annual observations on the German homeownership rate between 1998 and 2013; CBS provides seven observations for The Netherlands from 2006 to 2012. Further, Destatis and CBS provide statistics on the dwelling stock for Germany and The Netherlands, respectively. While data on the Dutch dwelling stock is available from 1947 to 2014, data on the German stock is only available from 1987 to 2013 (both stated annually). In addition, the OECD provides a house price-to-rent ratio index on a quarterly basis from 1970Q1 to 2013Q4 for both Germany and The Netherlands, which is used to proxy the cost of renting versus the cost of owning housing across both countries.

(30)

The German House Price Puzzle 24 The regulatory framework affecting the German or Dutch housing market is assessed on the basis of qualitative data (e.g., relating to the respective tax law) respectively. Hence, no quantitative data is obtained for this dimension.

To assess the countries’ housing supply market, statistics on issued building permits and the number of housing completions are obtained from Hypostat. The observations are stated on an annual basis and cover a sample period from 1996 to 2013 for both Germany and The Netherlands.

Moreover, Hypostat provides the number of transactions of new and second-hand dwellings in the housing market as a proxy for the investment market on an annual basis from 1996 to 2013, also for both countries.

Finally, the sixth dimension is denoted by other and predominantly focuses on assessing the nature of the German house price index. For the purpose of detecting a potential heterogeneity bias in the highly aggregated national German house price index, Bulwiengesa provides annual (nominal) house price data on all 16 federal states of Germany as well as on West and East Germany from 1990 to 2014. More explicitly, data is provided on five indicator variables comprising the average purchase price for single apartments per m2, rent level for new dwellings, rent level for second-hand dwellings, total purchase price for row houses, and purchase price for single-family property plots per m2, which are equally weighted to estimate the index series. The methodology is based on the Bulwiengesa house price index.

4.2 Descriptive Statistics

For the purpose of comparing the dynamics in house prices across selected housing markets, Table 5 provides estimates on the correlation between individual house price levels. The correlation matrix confirms the adverse price dynamics observed in the German housing market, visually indicated by the illustration in Figure 1. Over a sample period of 44 years, the level of German house prices is estimated to be strongly negatively correlated with international markets such as the US, the group of OECD countries as well as the UK or euro area. The correlation with

(31)

The German House Price Puzzle 25 the subject of comparison, The Netherlands, amounts negative -0.72, which represents a decrease in German house prices of -0.72% in case of a 1.00% increase in Dutch prices within the sample period. While the correlation with Germany ranges between -0.70 and -0.83, the correlation between other groups of countries is estimated to be strongly positive (e.g., 0.90 between The Netherlands and the euro area).

Table 5: Cross-Country House-Price Level Correlation Matrix

Notes: The table illustrates the correlation matrix between the OECD real house price index levels for Germany (DE), The Netherlands (NL), United Kingdom (UK), United States (US), euro area (EURO), and members of the Organisation for Economic Co-operation and Development (OECD). All Pearson’s correlation coefficients are significant at the 1% significance level based on 44 annual observations from 1970 to 2013.

In addition, Table 6 and Table 7 provide descriptive statistics on the level and changes in the employed macroeconomic variables, respectively. More explicitly, real house prices in Germany remained stable over the sample period from 1973 to 2012 relative to their value in 1970, observing a low volatility of 7.7%. While German prices peaked with a maximum of 12% (relative to prices in 1970), Dutch prices more than tripled, generally observing almost twice as high average prices and a significantly higher volatility. Less divergent, the German real disposable income per capita amounts €20,900, but is generally lower than in The Netherlands. Similarly, the nominal mortgage interest rate is generally lower in Germany and amounts on average 6.48% (7.23% in The Netherlands). Similarly, the real mortgage interest rates do not differ significantly across both countries, when accounting for the expected inflation. Finally, the real construction costs for new dwellings deflated by almost 50% in Germany over the sample period with an average of 71% relative to costs in 1970. In contrast, Dutch construction costs observed even more pronounced downward dynamics of up to 60% with an average of 53% relative to costs in 1970.

(32)

The German House Price Puzzle 26 Table 6: Descriptive Statistics on Employed Variables in Levels

Notes: The table illustrates selected descriptive statistics for the five principal macroeconomic fundamentals employed in the statistical analysis (including the nominal mortgage interest rate) comprising the mean, standard deviation (S.D.), minimum (low), and maximum (high) for both Germany and The Netherlands. The statistics are based on 40 annual observations from 1973 to 2012.

Assessing the changes in variables (i.e., first log differences) from 1973 to 2012, German real housing prices experience a minor negative trend, declining on average by approximately -0.3% per year (Table 7). More importantly, the growth rates of German house prices can be regarded as conservative and subdued, indicated by the low standard deviation of 2.6%. In stark contrast, Dutch house prices experience a strong positive trend of approximately 2.2% and a significantly higher volatility (over the same sample period, the standard deviation amounts 8.5%). When excluding the extraordinary volatility in the growth rates of Dutch house prices until 1982, the respective standard deviation between 1983 and 2012 still remains relatively high at 5.1% (2.4% for Germany). With respect to serial correlation, the changes in both German and Dutch house prices observe a highly significant first-order autocorrelation, however, with higher first-orders of lags being not significant.

Table 7: Descriptive Statistics on Employed Variables in First Differences

Notes: The table illustrates selected descriptive statistics for the five principal macroeconomic fundamentals employed in the statistical analysis (including the nominal mortgage interest rate) comprising the mean, standard deviation (S.D.), and the coefficient on the first-order autocorrelation (AC) for both Germany and The Netherlands. The statistics are based on 40 (39 for interest rates) annual observations from 1973 to 2012. All variables are stated in first-log differences. *, **, and *** indicate significance at the 10%, 5%, and 1% significance level, respectively.