Expectations-driven cycles in the housing market

Economic Modelling 60 (2017) 297–312

Contents lists available at ScienceDirect

Economic Modelling journal homepage: www.elsevier.com/locate/econmod

Expectations-driven cycles in the housing market a

c,⁎

b

Luisa Lambertini , Caterina Mendicino , Maria Teresa Punzi a b c

crossmark

EPFL, College of Management, Switzerland European Central Bank, Directorate General Research, Monetary Policy Research, Germany Vienna University of Economics and Business, Austria

A R T I C L E I N F O

A BS T RAC T

JEL classification: E32 E44 E52

This paper explores the transmission of “news shocks” in a model of the housing market and shows that anticipated signals or beliefs of future macroeconomic developments can generate boom-bust cycles in the housing market and lead to business cycle fluctuations. Anticipated monetary policy and inflationary shocks that turn out to be wrong can also lead to subsequent macroeconomic recessions. Credit frictions also play an important role in generating boom-bust cycle dynamics in the housing market. In particular, favorable credit conditions that are expected to be reversed in the near future generate an housing boom. The active use of the loan-to-value ratio as a policy tool aimed at dampening the severity of expectations-driven cycles effectively reduces the volatility of household debt, aggregate consumption and GDP.

Keywords: Boom-bust cycles Credit frictions Housing market

1. Introduction Boom-bust cycles in asset prices and economic activity are a central issue in policy and academic debates. Following the recent bursting of the housing bubble in the U.S. and the ensuing financial crisis, particular attention has been given to the behavior of housing prices and housing investment. This paper suggests a mechanism for modeling housing-market boom-bust cycles in accordance with the empirical pattern. Over the period 1965:1–2009:2 real house prices in the United States display a number of boom-bust episodes, namely periods of faster-than-trend growth followed by sharp reversals.1 See Fig. 1. Leamer (2015) documents that housing is the most critical part of the business cycle because, over 9 of 11 episodes' recessions since 1985, housing investment has declined three or four quarters before recessions.2 Indeed, over the last three decades, housing price boombust cycles in the United States have been characterized on average by hump-shaped co-movement in GDP, consumption, investment, hours worked, real wages and housing investment. More precisely, these macroeconomic variables generally grow during the boom phase of

housing prices and fall during the bust phase. Fig. 2 illustrates the behavior of a set of key macroeconomic variables during boom-bust episodes.3 Interestingly, as also highlighted by Leamer (2015), house prices and real residential investments peak several quarters before recessions, meaning that the housing market lead the business cycle. In particular, on average, the peak in house prices is anticipated by the peak in housing investment and followed by macroeconomic recessions.4 One possible interpretation is that the run-up in house prices and residential investments encourages household expenditure and households' loans. Once the demand for housing slow down, house prices start declining pushing towards an economic downturn. As a result a decline in house prices and worsened economic conditions, credit conditions also become tighter with further negative implications for housing and macroeconomic developments. Modeling endogenous boom-bust cycles in macroeconomics, however, is a major challenge. In business cycle models, it is difficult to generate extended periods of sustained house price growth followed by reversals through unanticipated shocks, which generate the strongest responses in the short run and eventually die out. An often-heard explanation of housing booms is households' optimism about future

⁎

Corresponding author. E-mail addresses: luisa.lambertini@epfl.ch (L. Lambertini), [email protected] (C. Mendicino), [email protected] (M.T. Punzi). 1 We define a peak as the centered maximum in real house prices in a twenty-one-quarter window. Using this definition we identify four boom-bust episodes that peaked in 1973:3; 1979:4; 1989:2; and 2006:2. 2 Leamer (2015) also documents the special case of housing downturn in 2006 in which housing peaked 2years before the start of the recession in 2008. Moreover, the recovery has been much delayed, relative to previous episodes. 3 All variables are an average of log-transformed, real, per capita around the four house price peaks. Appendix A describes the data in detail. Our results are robust to de-trending, either with a linear trend or with an Hodrick–Prescott filter. 4 Every housing peak as defined above has been followed by an economic downturn. Even the housing price high of 1969:4, which does not qualify as a peak according to our definition because real house prices rebounded too quickly, was followed by a recession. http://dx.doi.org/10.1016/j.econmod.2016.10.004 Received 23 January 2016; Received in revised form 26 September 2016; Accepted 9 October 2016 0264-9993/ © 2016 Elsevier B.V. All rights reserved.

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

related to consumers' belief of favorable buying conditions in the housing market both when the opinion is based on the perception of the current state of the economy and when it is driven by expectations of rising house prices or tighter future credit. Further, news heard of changes in business conditions contain statistically significant information for house price growth. These findings suggest a potential role for expectations-driven cycles in the housing market. Excess optimism about the housing market increases the demand for housing, leading to higher housing prices and to an excessively expansion of borrowing capacity. Mortgage financing builds up liquidity, which further encourages demand and residential investment, thereby contributing to the strong run-up in house prices, which in turn increases the collateral value that further releases more mortgage financing. This magnifies the housing boom phase and the co-movements with macroeconomic variables. In contrast, excess pessimism about the housing market generates sharp declines in housing and macroeconomic variables, leading to a deep recession. This is in line with the prospect theory (eg. Ahmad and Xiao, 2007) which states that possible gain and losses in wealth affect investment decisions of economic agents. Relative to gains, economic agents are much more hurt by losses in their wealth. Therefore business cycle fluctuations are a main concern because during recession, agents occur into losses of consumption. If investors are optimistic about the prospect of the economy, they would be willing to invest more and the asset prices would start to rise. On the other hand, pessimism about the economy will drag asset prices down.

Fig. 1. Real House Prices in the United States 1965:1–2009:2 (Data refers to new onefamily houses sold deflated with the implicit price deflator for the nonfarm business sector). The gray shaded areas indicate recession dates according to the National Bureau of Economic Research; the vertical line indicates the peaks in real house prices.

house price appreciation. Same authors document that beliefs of rising prices increased during the last housing boom (eg. Piazzesi and Schneider, 2009) and that changes in expectations of future house prices appreciation are important in predicting dynamics in house prices and other macroeconomic variables (eg. Lambertini et al., 2013; Huang, 2014). Using data from the Michigan Survey we document that news heard of recent changes in business conditions are significantly Real Consumption

Real Business Investment

Real GDP

0.45

0.39

0.44

0.38

0.75

0.70

0.37

0.43

0.36 0.65

0.42 0.35 0.41

0.34

0.40

0.60

0.33

0.39

0.55

0.32 2

4

6

8

10

12

14

16

18

2

20

4

6

8

Real Housing Investment

10

12

14

16

18

20

2

Real Loans

0.2

6

8

10

12

14

16

18

20

Hours Worked, Consumption Sector

0.75

0.3

4

0.16

0.15

0.70

0.14

0.1 0.65

0.13

0.0 0.60

-0.1

-0.2

0.12

0.11

0.55 2

4

6

8

10

12

14

16

18

20

2

4

6

8

10

12

14

16

18

20

2

4

Real Wage, Consumption Sector

Hours Worked, Housing Sector 0.15

8

10

12

14

16

18

20

18

20

Real Wage, Housing Sector

0.200

0.280

0.195

0.10

6

0.275

0.190 0.05

0.270 0.185

0.00

0.265

0.180

-0.05

0.175 2

4

6

8

10

12

14

16

18

20

0.260 2

4

6

8

10

12

14

16

18

20

2

4

6

8

10

12

14

Fig. 2. Macroeconomic variables average behavior during house-price boom-bust. The vertical line indicates the peaks in real house prices.

298

16

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

housing supply, the policy rate and the central banks target can generate housing-market boom-bust cycles characterized by co-movement in GDP, consumption, investment, hours and real wages. In contrast, expectations of a future increase in housing demand instead fail to generate boom bust cycles in the housing market in accordance with Figs. 1 and 2. According to Iacoviello and Neri (2010) housing demand shocks explain one-quarter of fluctuations in housing prices and housing investment in the United States over the last four decades. However, we document that expectations of a future increase in housing demand lead to a housing price boom but fail to generate co-movement between business investment and all other aggregate variables.6 In addition, only expectations of future expansionary monetary policy that are not met, both regarding the policy rate and the central banks inflation target, are likely to cause a boom-bust cycle and a macroeconomic recession. Thus, a high degree of transparency in monetary policy reduces uncertainty about future monetary policy actions and thereby the occurrence of cycles. Second, we also investigate the role of credit conditions as a source of housing market fluctuations. Survey data suggests that current favorable credit conditions as well as expected future tighter credit conditions are important reasons for consumers to judge house buying conditions as good. According to our findings, a current unanticipated increase in the access to credit raises on impact house prices and all other macroeconomic variables but fails to generate hump-shaped dynamics.7 We document that an easing of credit conditions generates boom-bust cycle dynamics in aggregate variables only if the current situation in the credit market is expected to be reversed in the near future. The fear of future worst credit conditions brings Borrowers to take advantage of the present loosening in credit standards, before the implementation of tighter financial conditions takes place. At a given higher mortgage demand, house prices increase and the financial accelerator mechanism further amplifies the housing boom. Last, we explore the role of credit conditions for the transmission of boom-bust cycles driven by news on shocks originated in other sectors of the economy. We find that lower Loan-to-Value ratios reduce the severity of boom-bust cycles in household debt, consumption and GDP. Accordingly, solving the model under a variety of unanticipated and news shocks, we find that lower LTV ratios imply lower volatility of these variables over the business cycle. These results highlight the importance of taking into account the effect of credit standards and financial innovations for house price dynamics into the standard macroeconomic frameworks used for policy analysis. The rest of the paper is organized as follows. Section 2 revises the previous literature. Section 3 explores the relationship between consumers' survey data and house prices growth. Section 4 describes the model. Section 5 investigates the occurrence of boom-bust cycles in the housing market as a consequence of expectations regarding future macroeconomic developments driven by news shocks. Section 6 shows the effect of current and expected credit conditions for boom-bust cycles and Section 7 investigates the role of credit conditions for macroeconomic fluctuations. Section 8 concludes.

We offer a model-based interpretation of the effects of news shocks on house prices and other macroeconomic variables. To this purpose, we explore the transmission mechanism of news shocks in a dynamic stochastic general equilibrium (DSGE) model of the housing market. We show that news shocks can generate boom and busts dynamics in the housing market and other macroeconomic variables. We develop this argument by augmenting the model of Iacoviello and Neri (2010) to include news shocks. Their framework features a certain degree of granularity that allows us to explore the role of news on a variety of (both real and nominal) underlying sources of fluctuations, such as productivity, housing demand and supply, monetary policy and credit market conditions. The model features two type of households (ie. Savers and Borrowers), three production sectors (ie. production of consumption goods, capital and housing), collateralized household debt, and a monetary authority that follows a Taylor-type rule. Iacoviello and Neri (2010)'s model belongs to the recent literature that documents the role of collateral constraints for housing market fluctuations.5 This class of models assumes the existence of ex ante heterogeneous households who differ in terms of their discount factors: patient agents desire to accumulate savings, while impatient agents prefer to consume most of their wealth. This provides a motive for the existence of credit flows in the model. As common in the credit frictions literature, households can borrow against the value of their houses. In addition, both the demand and supply of housing are modeled and the house price is endogenously determined. According to Iacoviello and Neri (2010)'s findings housing preference shocks, housing technology shocks and monetary shocks are the main driving sources of businesscycle fluctuations and help replicating the correlations between GDP, house prices, residential and non-residential investments, households' indebtedness and real wages. However, their analysis does not aim at replicating the hump-shaped behavior of macroeconomic and housing variables observed during periods of housing booms. Our paper fills this gap and shows that it is possible to replicate the empirical pattern documented in Figs. 1 and 2 once news shocks are taken into account. Our main findings as summarized as follows: (i) we provide evidences of the role of expectations of future developments in several sectors of the economy to generate housing-market boom-bust cycles characterized by co-movement in GDP, consumption, investment, hours and real wages, (ii) we emphasize the fact that exogenous easing of credit conditions that are expected to be reversed in the near future also generates boom-bust cycle in the housing market; (iii) we investigate the effectiveness of lowering the Loan-to-Value (LTV) ratios in reducing the severity of boom-bust cycles in household debt, consumption and GDP. Our paper analyzes whether news on relevant sources of macroeconomic fluctuations are able to generate expectations-driven cycles in the housing market. This paper provides several insightful results. First, regarding the transmission mechanism of news shocks the transmission of news shocks to the housing market is linked to the well-known financial accelerator mechanism. Boom-bust cycles emerge if agents expect a future increase in housing prices, which fuels current housing demand and lifts housing prices immediately. Thus, the increase in housing prices is coupled with an endogenous increase in household indebtedness. Similar to Christensen et al. (2010), we find that price and wage stickiness have an important role for expectations on future productivity to generate co-movement between house prices and consumption, investment and hours worked. We obtain boom-bust dynamics in all aggregate variables and real wages, whereas Christensen et al. (2010) fail to generate a co-movement with real wages. We document that not only changes in expectations about future productivity, but also news on a variety of other shocks, such as

2. Related literature According to Beaudry and Portier (2006) business cycle fluctuations in the data are primarily driven by changes in agents' expectations about future technological growth. In fact, they first document that 6 This result is in line with Liu et al. (2013) that document that in the absence of credit frictions at the firm level housing demand shocks cannot explain the dynamics of business investment. 7 This result is in line with previous works that fail to replicate an housing boom-bust cycle due to an unanticipated credit shock, ie. an innovation in the LTV ratio. In this respect, Iacoviello and Neri (2010) and Mendicino and Punzi (2014) find that an unanticipated credit shock fails to replicate the co-movements of residential and business investments.

5 See also, Campbell and Hercowitz (2004), Iacoviello and Minetti (2006), Liu et al. (2013), Finocchiaro and Heideken (2013), Walentin (2014a,b), Christensen et al. (2016), and Calza et al. (2013).

299

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

role for expectations on future productivity to generate co-movement between house prices and consumption, investment and hours worked. However, we obtain boom-bust dynamics in all aggregate variables and real wages, whereas Christensen et al. (2010) fail to generate a comovement with real wages.10 Compared to the above-mentioned literature, we also show that other type of news shocks generates boom-bust cycle dynamics in house prices and other macroeconomic variables. Only few other papers relate booms and busts in the housing market to expectations on future fundamentals, ie. anticipated news shocks. Tomura (2010) documents that boom-bust cycles in house prices can be generated by uncertainty about the duration of a temporary increase in income growth only if the economy is open to international capital flows. Relative to Tomura (2010), we analyze the transmission mechanism not only of sudden signal about future technological progress, but also news on Central Banks' actions and news on other sectors of the economy. Adam et al. (2012) explain the joint dynamics of house prices and the current account over the years 2001–2008 by relying on a model of “internally rational” agents that form beliefs about how house prices relate to economic fundamentals; Burnside et al. (2012) generate boom and bust in the housing market relying on heterogeneous expectations about long-run fundamental that drive house prices, as summarized by the flow of utility of holding a house. Kuang (2014) shows that distorted beliefs amplify house price cycle because surprise in house prices fuels optimism and credit expansion, by raising housing demand. Such optimism reinforces agents' price beliefs and further expands credit. Different from these other papers of the housing market, we do not relate to just one type of news shock, ie. news on productivity. Following recent developments in the news shock literature, we explore the transmission mechanism of news on a variety of other shocks (see Schmitt-Grohe and Uribe, 2012; Khan and Tsoukalas, 2012; Milani and Treadwell, 2012; Gomes et al., 2014). In comparison to previous DSGE analysis of the housing market, we show that only through news shocks it is possible to replicate the comovement among housing and macro variables observed during periods of periods of housing booms and busts. Indeed, even the unanticipated sources of business cycle dynamics highlighted in the credit friction literature, ie. Iacoviello (2005), Iacoviello and Neri (2010) and Liu et al. (2013), fail to generate the hump-shaped behavior of macroeconomic and housing variables. It is important to stress that several papers carry out a quantitative analysis of housing-market dynamics. However, only a few aim at explaining business-cycle fluctuations in both house prices and investment.11 The framework developed by Iacoviello and Neri (2010) fits the purpose of this paper better compared to other models, since it allows us to explore the role of news on a variety of shocks, including monetary policy and credit market shocks.12 Recent papers study the policy implications of

stock prices movements anticipate future growth in total factor productivity and that such dynamics are accompanied by a macroeconomic boom. However, as already shown by Beaudry and Portier (2004, 2007), a standard one-sector optimal growth model is unable to generate boom-bust cycles in response to news. At the time of the signal consumption increases and hours worked fall thanks to the wealth effect generated by expectations of improved future macroeconomic conditions. Since technology has not improved yet, output decreases. In order for consumption to increase despite the reduction in hours worked, investment has to fall. Thus, good news creates a boom in private consumption and a decline in hours worked, investment and output. A two-sector model with consumption and capital goods is also unable to generate a boom in macroeconomic variables.8 When a three-sector model is considered, Beaudry and Portier (2004, 2007)) show that expectations-driven cycles can arise provided firms exhibit economy of scope or, in other words, internal cost complementarities between the production of different goods. Jaimovich and Rebelo (2009) introduce three elements in an otherwise standard neoclassical growth model: Variable capital utilization; adjustment costs to investment; and a weak short-run wealth elasticity of labor supply. This latter element is introduced by assuming a generalized version of the preference specification considered by Greenwood et al. (1988). A one-sector model displays co-movement of consumption, output, investment and hours worked in response to news about future total factor productivity or about investment-specific technology. The value of the firm, however, falls unless the production function features decreasing returns to scale as stemming from a factor of production in fixed supply. A two sector model is able to generate comovement in response to news about future aggregate productivity, productivity in the consumption sector, and productivity in the investment sector only provided the short-run wealth effects on the labor supply are very low, the elasticity of labor supply is high and the elasticity of capacity utilization is low. Jaimovich and Rebelo (2009) also explore a version of their two-sector model with adjustment costs to labor and find that they are helpful to generate co-movement in response to news.9 Christensen et al. (2010) show that a standard one-sector real business cycle model with habit persistence and costs of adjusting the flow of investment generates a boom-bust pattern in output, consumption, investment and hours in response to news on productivity that do not materialize. The price of capital, however, is negatively correlated with all other aggregate variables and therefore it falls and then increases. The introduction of an inflation targeting central bank and sticky nominal wages make the price of capital co-move with the other aggregate variables and boom-bust dynamics emerge. When news spreads about a future increase in productivity, aggregate variables increase including hours worked. The increase in hours is possible because the real wage falls, hence producers are willing to raise labor demand. Since nominal wages are sticky, a decrease in real wages occurs because prices increase faster than wages. This paper shows that news on future productivity shocks generate macroeconomic booms also in a model of the housing market that features collateralized household debt, standard preferences and production functions and nominal rigidities. Similar to Christensen et al. (2010), we find that price and wage stickiness have an important

10 Intuitively, the increase in housing demand and therefore housing prices in response to news allows for an increase in both real wages and hours in the housing sector that spills over the consumption sector. In contrast, in Christensen et al. (2010) real wages fall and remain below trend for most of the boom phase and then increase shortly before the period when the shock is going to be realized (or not). The empirical evidence in Appendix B seems to suggest that real wages are not below trend before a peak in house prices and that they increase throughout the boom phase. Notice also that the asset-price peak in the third quarter of the year 2000 to which Christensen et al. (2010) refer to was preceded by rapid growth of real wages both in the consumption-good and in the housing sector and real wages were also above trend. 11 Among these, Davis and Heathcote (2005) use a calibrated multi-sector model that relies on technology shocks to match the co-movement between consumption, nonresidential investment, residential investment and GDP; Int'Veld et al. (2011) build an open-economy model of the housing market, also featuring an analysis of the banking sector to explore the international repercussions of housing-market dynamics. 12 An alternative model describing the feedback mechanism of house prices to the macroeconomy can be found in Adams and Füss (2010) that show that higher economic activity and construction costs lead to an increase in house prices, while an increasing long-term interest rate lowers house prices. Their analysis, however, abstracts from the

8 News about productivity in the capital sector raises consumption but reduces hours worked. As a result, investment, capital and output fall. An announcement of future higher productivity in the consumption sector generates a boom in all macro variables except consumption for elasticities of intertemporal substitution above one; vice versa, it generates a bust in all macro variables but consumption when the elasticity is below one. 9 Other papers have focused on different mechanisms. Den Haan and Kaltenbrunner (2009) consider a labor market matching mechanism; Floden (2007) incorporates variable capital utilization and vintage capital; Kobayashi et al. (2007) and Walentin (2014a) shows that expectations-driven cycles can arise in models with credit constraints on firms; Nutahara (2010) proves that in contrast to external habits, internal habits can help to generate co-movement in response to news on future productivity.

300

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

housing market boom bust cycle driven by news shocks (eg. Lambertini et al., 2013; Kanik and Xiao, 2014) in a similar framework. This paper complements the analysis in Lambertini et al. (2013) by providing a better understanding of the transmission mechanism of news in generating housing boom-bust cycles. We also highlight the importance of credit conditions as a source of boom-bust cycles in the housing market. In addition, while Lambertini et al. (2013) discuss the implications of a time-varying Loan-to-Value ratios, revealing an important trade-off in terms of welfare achievement from different types of agents, we discuss the implications of policies with the ultimate goal of reducing volatility.

Table 1 Correlation between News on Business Conditions and Good Time to Buy. Variables

Full sample

Boom 2006

Good Time to Buy, Prices Low

−0.2160*** [0.0036] 0.2686*** [0.0003] 0.2847*** [0.0001] 0.1454** [0.0514]

−0.7649*** [0.0000] 0.5830*** [0.0055] 0.6298*** [0.0022] 0.7008*** [0.0004]

Good Time to Buy, Prices Rising Good Time to Buy, Future Tighter Credit Good Time to Buy, Credit Easy

Full sample: 1965:1–2009:4. P-values in parentheses. * 10% significance. ** 5% significance. *** 1% significance.

3. Survey data and house market dynamics This section provides some suggestive evidence on the importance of news for housing market dynamics. We rely on survey data from the University of Michigan Survey of Consumers, which provides assessments of consumer attitudes and expectations. We focus on two questions: (a) News Heard of Recent Changes in Business Conditions and (b) Buying Conditions for Houses. The index of News Heard of Recent Changes in Business Conditions (News on Business Conditions henceforth) reports the fraction of respondents who heard favorable news minus the fraction of respondents who heard negative news of recent changes in business conditions.13 Appendix A reports the exact question wording for this variable. As for Buying Conditions for Houses, the survey reports the consumers' opinion as to whether it is a good time or a bad time to buy a house and their reasons for holding a particular view. The possible reasons for opinions about good or bad buying conditions for houses can be summarized in the following categories: low house prices, higher future house prices, low current interest rates, and tighter future borrowing conditions. The variable reports the fraction of respondents citing that specific reason – see Appendix A for details. We construct four variables using this data. The first variable is Good Time to Buy-Prices Low, which we calculate as the difference between Good time to buy (house prices are low) and Bad time to buy (house prices are high); the second variable is Good time to Buy-Prices Rising, which is equal to Good time to buy (house prices will not come down; are going higher); the third variable is Good time to Buy-Future Tighter Credit, which is equal to Good time to buy (borrow-in-advance of rising interest rates); and the last variable is Good time to Buy-Credit Easy, which is calculated as the difference between Good time to buy (interest rates are low; credit is easy) and Bad time to buy (interest rates are high; credit is tight).14 We analyze the correlation between the index of News on Business Conditions and the four variables reporting good buying conditions in the housing market. Table 1 shows that News on Business Conditions is positively and significantly correlated with Good Time to Buy because of current low interest rates or because consumers expect housing prices to increase or credit conditions to tighten in the future. During

the latest boom episode these correlations become even higher. Interestingly, News on Business Conditions is negatively correlated with Good Time to Buy because housing prices are low and this correlation is very strong during the last boom episode. To assess causality, we run a Granger causality test, whose results are reported in Table 2. The index of News on Business Conditions contains statistical significant information for consumers' perception of Good Time to Buy both when the opinion is driven by perceptions on the current state of the economy related to low housing prices and easy credit and when the opinion is based on expectations of tighter future credit or of future house prices appreciation. Hence, News on Business Conditions Granger cause consumers' perception that is it is a Good Time to Buy a house.15 Table 3 shows that the correlation of News on Business Conditions with house price growth is high and significant over the full sample and typically higher during house price boom episodes. Moreover, News on Business Conditions contains statistical significant information for house price growth. In fact, the hypothesis that News on Business Conditions does not Granger cause house price growth can be rejected at the one per cent significance level. On the contrary, house price growth do not contain significant information to explain the fact that consumers heard news of changes in business conditions. See Table 4. Summarizing, both consumers' assessments of buying conditions for houses and housing prices are explained by past values news on business conditions. Moreover, news on business conditions, expectations of future macroeconomic developments and economic optimism, are significantly related to house price growth.16 4. The model economy We adopt the framework developed by Iacoviello and Neri (2010) that introduces financial frictions at the household sector in an otherwise standard two-sector dynamic stochastic general equilibrium (DSGE) model of the housing market. The model considers two types of agents, ie. Savers and Borrowers, and financial frictions are introduced in the form of collateral constraints: Borrowers can only raise credit up to a certain fraction of their housing value. This helps to match the positive correlation between housing prices and investment and the wealth effect of housing prices. In addition, the model features nominal price and wage rigidities and a larger set of shocks. We extend the original model to also include news shocks. the framework developed by Iacoviello and Neri (2010) features the right degree of granularity that allows us to explore the role of news on a variety of (both real and nominal)underlying sources of fluctuations such as productivity in the consumption-, capital- and housing sector, housing demand and

(footnote continued) modeling of heterogeneous household, the importance of the mortgage market through the collateral constraint, the labor market in the good and construction sectors and the importance of monetary policy through the Taylor rule. 13 Several authors have used the Surveys of Consumers of the University of Michigan to document the fact that expectations of future house price increases play an important role in periods of rising house prices. Piazzesi and Schneider (2009) using the survey data on Expectations of Rising House Prices as a proxy for house price expectations find that beliefs that house prices will rise, as measured by the percentage of agents who believe that prices will rise, increased in the United States during the last housing boom. In addition, they argue that expectations of future house price appreciation are related to optimism about economic conditions. See, among others, also Lambertini et al. (2013) and Huang (2014). 14 Consumers appear to assess home buying conditions quite well. In fact, changes in home buying attitudes precede changes in unit sales of new and existing single family homes on average by two quarters with a correlation of 0.77.

15 Lag has been selected using both Schwarz information criterion and Hannan– Quinn information criterion. 16 News media data have also been used to measure optimism and housing sentiment. See, eg. Soo (2013), Walker (2014), Nofsinger (2012) and Foote et al. (2012).

301

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

late housing, h t . They differ in their discount factor ( β and β′). ′ Borrowers (denoted by ′) feature a relatively lower subjective discount factor that in equilibrium generates an incentive to anticipate future consumption in the current period through borrowing. Hence, the ex ante heterogeneity induces credit flows between the two types of agents. This modeling feature has been introduced in macro models by Kiyotaki and Moore (1997) and extended by Iacoviello (2005) to a business cycle framework with housing investment. The Borrowers: The Borrowers maximize the utility function17:

Table 2 Granger causality tests between News on Business Conditions and Good Time to Buy. Null hypothesis

F-statistic

News on Business Conditions does not Granger Cause GTB, Prices Low

10.9246*** [0.0000] 0.6911

GTB, Prices Low does not Granger Cause News on Business Conditions

[0.5024] News on Business Conditions does not Granger Cause GTB, Prices Rising GTB, Prices Rising does not Granger Cause News on Business Conditions

∞

3.7857***

Ut = Et

t =0

[0.0246] 0.8791

⎡

∑ β′t GCt ⎢ln(ct′ − ε′ct′−1) + j ln ht′ ⎣

1+ η ′ ⎤ τ − ((nc′, t )1+ ξ ′ + (nh′, t )1+ ξ ′) 1+ ξ ′ ⎥ ⎦ 1 + η′

[0.417] News on Business Conditions does not Granger Cause GTB, Future Tighter Credit GTB, Future Tighter Credit does not Granger Cause News on Business Conditions

subject to the budget constraint:

5.3644***

ct′ + qt [ht′ − (1 − δh ) ht′−1] − bt′ ≤

[0.0055] 0.6210

We allow Borrowers to collateralize the value of their homes:

[0.5386] News on Business Conditions does not Granger Cause GTB, Credit Easy GTB, Credit Easy does not Granger Cause News on Business Conditions

bt′ ≤ mEt

[0.0183] 5.5639

Except for the gross nominal interest rate, R, all the variables are expressed in real terms; πt is the gross inflation (Pt / Pt−1), w′c, t and w′h, t are the wages paid in the two sectors of production, and qt is the price of housing in real terms. Houses depreciate at rate δh and j determines the relative weight in utility on housing services and it represents a proxy for housing demand shocks. Limits on borrowing are introduced through the assumption that households cannot borrow more than a fraction of the next-period value of the housing stock. See Eq. (1). The fraction m, referred to as the equity requirement or loan-to-value ratio, is treated as exogenous to the model. It can be interpreted as the creditor's overall judicial costs in the case of debtor default and represents the economy's degree of access to the credit market. The borrowing constraint is consistent with standard lending criteria used in the mortgage and consumer loan markets. The Savers: The Savers face a similar problem to that of the Borrowers in choosing how much to consume, how much to work, and how much to invest in housing. However, they also invest in capital and receive the profits of the firms.18 Firms: Final good producing firms produce non-durable goods (Y) and new houses (IH). Both sectors face Cobb–Douglas production functions. The non-housing sector produces consumption goods using capital, kc, and labor supplied by the Savers, n, and the Borrowers, n′,

2 lags; sample: 1965:1 to 2009:4. GTB: Good Time to Buy. * 10% significance. ** 5% significance. *** 1% significance. Table 3 Correlation between House Price Growth and News on Business Conditions.

***

0.6857 [0.0000]

Boom 1973 ***

0.7451 [0.0001]

Boom 1979 ***

0.6854 [0.0006]

Boom 1989 ***

0.7268 [0.0002]

Boom 2006 0.7389*** [0.0001]

Sample: 1965:1–2009:4; P-values in parentheses. 10% significance. 5% significance. *** 1% significance.

*

**

Table 4 Granger causality tests between Housing Price and News on Business Conditions. Null hypothesis

2 lags F-statistic

News on Business Conditions does not Granger Cause House Price Growth

19.0776***

House Price Growth does not Granger Cause News on Business Conditions

qt +1 πt +1 h t ′ . Rt

4.0961***

[0.0046]

Full sample

wc′, t nc′, t wh′, t nh′, t R b′ + − t −1 t −1 ′ ,t ′ ,t Xwc Xwh πt

1− μc

α Yt = (Ac, t (ncα, t nc′,1− t ))

(1)

(z c, t kc, t −1) μc .

The housing sector also uses an intermediate input, kb, and land, l, as inputs of production

[0.0000] 2.66938***

1− μh − μ − μ b l (z

IHt = (Ah, t (nhα, t + nh′, t1− α ))

μ μb μl h, t kh, t −1) h kb lt −1.

[0.0722] 2 lags; sample: 1965:1–2009:4. * 10% significance. ** 5% significance. *** 1% significance.

17 Since consumption grows at the rate GC every quarter, the marginal utility of consumption falls at the same rate. For this reason, the model is transformed so that consumption is scaled by its growth rate GtC and the marginal utility of consumption is multiplied by the same rate. As a result, transformed consumption and scaled marginal utility of consumption are both constant in the steady state. 18 The model we rely on does not allow for a convex adjustment cost of housing demand. Iacoviello and Neri (2010) report that in preliminary estimation attempts, they found that the parameter measuring this cost was driven to its lower bound of zero. As argued by Iacoviello and Neri (2010), 25 percent of residential investment in the National Income and Product Accounts (that is the variable used in the data analysis) consists of home improvements, where transaction costs are less likely to apply. However, we acknowledge that the existence of substantial transactions costs (both pecuniary and search) creates a nontrivial possibility for serial correlation in excess return that can affect the formation of house price expectations and the shapes of the responses of house prices to shocks.

supply, monetary policy and credit market conditions. The models parameters are set equal to the mean of the posterior distribution estimated by Iacoviello and Neri (2010) for the U.S. economy. For completeness we report the main features of the model in the following: Households: The economy is populated by two types of households: the Saver and the Borrower. They both work in the production of consumption goods, n′c, t , and housing, n′h, t , consume, nt′, and accumu302

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

The investment-specific technology is as follows:

Table 5 Calibrated parameters. Source: Iacoviello and Neri (2010).

1 IKt = (kc, t − (1 − δ kc ) kc, t −1). Ak , t Ac, t represents a standard aggregate technology shock, Ak , t is an investment-specific shock and Ah, t is a specific productivity shock in the housing sectors. The model assumes heterogeneous deterministic trends in all these shocks. Firms pay the wages to households and repay the rented capital to the Savers. Retailers, owned by the Savers, differentiate final goods and act in a competitive monopolistic market. Prices can be adjusted with probability 1 − θπ every period, by following a Calvo-setting. Monopolistic competition occurs at the retail level, leading to the following forward-looking Phillips curve: ln πt − ιπ ln πt −1 = β (Et ln πt +1 − ιπ ln πt ) − ϵπ ln(Xt / X ) (1 − θ )(1 − βθ )

π π where ϵπ = , and Xt represents the price markup.19 θπ Households set wages in a monopolistic way. Wages can be adjusted subject to a Calvo scheme with a given probability in every period. Housing prices are assumed to be flexible. Monetary authority: We assume that the central bank follows a Taylor-type rule as estimated by Iacoviello and Neri (2010):

⎛ GDPt ⎞(1− rR) rY uR, t R π (1− rR ) rπ Rt = Rtr−1 rr (1− rR) ⎜ ⎟ t ⎝ GDPt −1 ⎠ st

Value

Parameter

Value

Parameter

Value

β β′ j

0.9925 0.97 0.12

ξ ξ′ ϕk , c

0.66 0.97 14.25

ρs ρAC ρAH

0.975 0.95 0.997

μc

0.35

ϕk , h

10.90

ρAK

0.992

μh μl μb δh δkc δkh X Xwc Xwh m ε ε′ η η′

0.10 0.10 0.10 0.01 0.025 0.03 1.15 1.15 1.15 0.85 0.32 0.58 0.52 0.51

α rR rπ rY θπ ιπ θ w, c ιw, c θ w, h ιw, h ζ γAC γAH γAK

0.79 0.59 1.44 0.52 0.83 0.69 0.79 0.08 0.91 0.40 0.69 0.0032 0.0008 0.0027

ρj ρz ρτ σAC σAH σAK σj σR σz στ σp σs σ w, c σ w, h

0.96 0.96 0.92 0.0100 0.0193 0.0104 0.0416 0.0034 0.0178 0.0254 0.0046 0.0004 0.1218 0.0071

Agents' endogenous expectations could be related to both current and expected macroeconomic developments. Thus, unanticipated and news shocks are both potential sources of agents expectations on economic conditions and house prices. However, business cycle models that feature only standard unanticipated shocks as sources of fluctuations cannot reproduce the hump-shaped dynamics shown in the data during periods of boom-bust in the housing market. In fact, standard unanticipated shocks generate the strongest responses in the short run and eventually die away. Moreover, in the current framework, macroeconomic developments led by unanticipated shocks also fail in generating the observed co-movement of housing prices with hours worked, investment and GDP. See Appendix C.

(2)

where rr is the steady state real interest rate and uR, t is an i.i.d. monetary policy shock. The central bank's target is assumed to be time varying and subject to an AR(1) shock, st:

st = (1 − ρs ) st −1 + us, t .

Parameter

(3)

GDP is defined as the sum of consumption and investment at constant prices. Thus

GDPt = Ct + IKt + qIHt , where q is the real housing prices at the steady state. The model's parameters are set equal to the mean of the posterior distribution estimated by Iacoviello and Neri (2010) for the U.S. economy. See Table 5.

5.1. News on productivity Changes in agents' expectations about future productivity seem to be an important source of business cycle fluctuations, see eg. Beaudry and Portier (2006). In the following, we show that news on future productivity shocks can generate a housing market boom. Fig. 3 reports the effect of an anticipated increase in productivity, namely a fourperiod-ahead expected shock to Ac, t (starred line). It also illustrates the case in which news of a future shock to Ac, t turns out to be wrong, and at time t=4 there is no change in productivity (solid line). Expectations of future productivity gains generate boom-bust dynamics in GDP, consumption, hours, investment and house prices. The intuition is as follows. The expected wealth effect caused by positive future economic conditions that bring households is to increase their current consumption expenditure as the value of their assets is expected to increase. Due to demand pressures, inflation increases. At the same time, the anticipation of higher future productivity also generates expectations of increasing housing prices. Higher expected housing prices lead to an increase in Borrowers' housing expenditure and indebtedness. The increase in current house prices is mainly driven by the increase in the housing demand by Borrowers, who can use the value of the house as a collateral for external financing. In contrast, Savers' housing demand declines, since they prefer to invest in business opportunities, including lending to Borrowers given the higher real interest rate. Moreover, due to limits to credit, impatient households increase their labor supply in order to raise internal funds for housing investment. Adjustment costs of capital also play an important role in the

5. News shocks News shocks can result from announcements of an increase or a decline in future productivity or credible central bank announcements about implementing interest rate paths that deviate from the usual pattern, as captured by the systematic part of a Taylor-type rule. Alternatively, they may result from the private sector's own beliefs about future unanticipated macroeconomic developments. To introduce expectations of economic developments, we follow Christensen et al. (2010) in assuming that the error term of each shock consists of an unanticipated component, εz, t , and an anticipated change n quarters in advance, εz, t − n ,

uz, t = εz, t + εz, t − n , where εz, t is i.i.d. and z = {Ac , Ak , Ah , π , R}. Thus, at time t agents receive a signal about future macroeconomic conditions at time t + n . If the expected movement does not occur, then εz, t = −εz, t − n and uz, t = 0 .20 19 The model allows for price rigidities in the consumption sector and for wage rigidities in both the consumption and housing sectors. There are several reasons why housing might have flexible prices. For instance, Barsky et al. (2007) argue that housing is relatively expensive on a per-unit basis. Therefore, if menu costs have important fixed components, there is a large incentive to negotiate on the price of this good. Moreover, most homes are priced for the first time when they are sold. 20 Schmitt-Grohe and Uribe (2012) assume that the error term is equal to εx, t = εx, t + εx, t −4 + εx, t −8 , however their estimation shows that the anticipated component of 8-quarters ahead explains almost nothing of the variance decomposition of GDP, consumption and business investment. Therefore, we keep a simple specification of the anticipated component of the shock with 4-quarters ahead. The estimation and

(footnote continued) importance of anticipation components are beyond the purpose of this paper.

303

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

Fig. 3. News on technology shock: anticipated shock (starred line) and not realized shock (solid line).

mechanism transmission of news to the macroeconomy because firms in the consumption sector already start adjusting the stock of capital at the time in which news spreads about a future increase in productivity. This way, when the increase in productivity occurs, capital is already in place. For the increase in business investment to be coupled with an increase in total hours worked, wages must rise. The increase in business and housing investment makes GDP rise. In the case of an anticipated shock that is realized, aggregate variables boom and then slowly decline (starred line). The peak response in output corresponds to the time at which expectations realize. In contrast, if expectations do not realize there is a dramatic drop in both quantities and prices.21 The model presented above features several real and nominal rigidities. In order to disentangle the contribution of the different modeling choices, we introduce the frictions one at the time. Fig. 4 displays the boom-bust response to news on productivity in the flexible-price version of the model coupled with (a) the absence of adjustment costs of capital and inability of impatient households to borrow (dashed line), ie. when m=0, (b) adjustment costs of capital and inability of impatient households to borrow (starred line) and (c)

adjustment costs of capital and ability of impatient households to borrow (solid line). Under the (a) scenario, the absence of price stickiness generates a strong wealth effect agents increase in both consumption and leisure. To increase consumption households reduce their investment expenditures (in all sectors), leading to a decreasing GDP. The weak demand pressure generates disinflation and decreasing real interest rate, discouraging the Savers to lend to Borrowers. As a result, house prices decrease due to lower demand for mortgage loans. Under the (b) scenario, when adjusting capital is costly, the reduction in business investment and thus the increase in consumption is less pronounced (starred line). However, the inability of Borrowers to increase their housing demand because of lack of housing financing prevent the run-up in house prices. Under scenario (c), allowing for borrowing against the value of collateral leads to a more pronounced increase in Borrower's housing demand (solid line). In this last case, Borrower's consumption increases by more in the boom phase and the decline in Borrower's hours (not shown in the graph) is more sizable. Saver's demand for housing declines. Since Savers account for about eighty percent of labor income, aggregate housing production declines due to a lower supply of labor from Saver in the construction sector, leading to a fall in house prices. To sum up, adjustment costs and the collateral effect are not enough to generate boom-bust dynamics in the absence of nominal rigidities. Fig. 5 shows the response of the economy with nominal rigidity in the price of the consumption good but no wage rigidities (dashed line).

21 Similar to news on productivity shocks, expectations of a future increase in the cost of transforming output into capital, Ak , generate an increase in the production of consumption goods, housing investment and business investment, GDP. However, unrealized expectations induce a faster return to the initial state, without implying below-steady state dynamics. See Appendix D.

304

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

Fig. 4. News on technology shock – flexible-price case.

tially, thereby reducing labor demand in the sector. To obtain a boom in investment and hours in the housing sector it is necessary to introduce wage stickiness in the housing sector. Finally, we add wage stickiness in both sectors of production (solid line). Since wage stickiness is more sizable in the housing sector, the increase in wage in that sector is less pronounced. Due to a further reduction in the current income effect, agents increase their labor supply by more. Aggregate housing investment increases more so that housing prices rise less. Household debt increases less but aggregate consumption is barely affected relative to the case of no wage stickiness in the housing sector. Thus, in the presence of nominal price and wage rigidities, expectations of future productivity gains generate empirically plausible boom-bust cycle dynamics. As in Christensen et al. (2010) we show that price and wage stickiness have an important role for expectations on future productivity to generate co-movement between house prices and consumption, investment and hours worked. However, contrary to them, we obtain boom-bust dynamics in all aggregate variables and real wages. In our model house prices co-move with the other aggregate variables independent of whether wages are stickier than prices or vice versa. Intuitively, the increase in housing demand and therefore housing prices in response to news allows for an increase in both real wages and

Expectations of higher future productivity lead to a decrease in expected inflation, which in turn reduces the expected real interest rate. The decline in the current real interest rate coupled with a higher expected real rate leads to an increase in current debt and thus Borrowers' consumption, Borrowers' housing demand and Savers' consumption. On the contrary, Savers reduce their housing demand and increase their supply of labor. For a contemporaneous increase in business investment and hours, the rise in wages in the consumption sector needs to be significant. Aggregate housing investment first declines and then slowly increases; housing prices increase as well as current inflation. However, compared to the case with flexible prices, inflation rises by less, thereby allowing for a more pronounced increase in consumption. In the additional presence of wage stickiness in the consumption sector, the wage in the consumption sector increases by less (starred line), which raises the demand for labor and therefore hours in the consumption sector. Moreover, since the sectorial wage differential is more pronounced, Savers increase their labor supply in the housing sector as well. Thus, the model displays co-movement of GDP, consumption, business investment and housing prices over the boom-bust cycle. Housing investment and hours in the housing sector, however, fall because wages in the housing sector increase substan-

305

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

Fig. 5. News on technology shock – stickiness case.

hours in the housing sector that spills over the consumption sector.22

technology news, with wage-markup news shocks in particular accounting for about sixty per cent of the variance share of both hours and inflation. A few authors also argue that anticipated monetary policy shocks play a larger role in the business cycle than unanticipated shocks (eg. Milani and Treadwell, 2012; Gomes et al., 2014). Since boom-bust cycles in the housing market can be plausibly related to expectations of future developments in different sectors of the economy, in the following, we explore the effects of changes in expectations on several other shocks.23 In particular, we show that changes in expectations about future housing market shocks, investment-specific shocks and monetary policy shocks can also generate housing-market boom-bust cycles characterized by co-movement in GDP, consumption, investment, hours and real wages.

5.2. News on other shocks The empirical literature on news shocks is growing rapidly. Several papers document that news on a variety of other shocks are important sources of business cycle fluctuations. Among others, Schmitt-Grohe and Uribe (2012) show that innovations in expectations of future neutral productivity shocks, investment-specific shocks, and government spending shocks account for more than two-thirds of predicted aggregate fluctuations in postwar United States. Khan and Tsoukalas (2012) show that, in the presence of wage and price rigidities and a variety of news shocks, non-technology sources of news dominate

5.2.1. News on housing demand and supply Iacoviello and Neri (2010) document that housing demand and

22 The empirical evidence in Fig. 14 seems to suggest that real wages are not below trend before a peak in house prices and that they increase throughout the boom phase. Notice also that the asset-price peak in the first quarter of the year 2000–2001 to which Christensen et al. (2010) refer to was preceded by a rapid increase in real wages both in the consumption-good and in the housing sector – see Fig. 9 in the Appendix. In their model, the increase in hours is possible because the real wage falls, hence producers are willing to raise labor demand. Since nominal wages are sticky, a decrease in real wages occurs because prices fall faster than wages. The inflation-targeting central bank responds to this fall in inflation by cutting the nominal interest rate, which in turn raises investment and the price of capital.

23 Duca et al. (2010) highlight that large swings in housing construction have had major macroeconomic effects in Ireland, Spain, and the US. Bjørnland and Jacobsen (2010) found that unexpected changes in interest rates have an immediate effect on house prices in most countries. They found that monetary policy contributes significantly to house price fluctuations, and that house price innovations are, in turn, important for variability in macro variables.

306

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

Fig. 6. Unrealized News Shocks on: Housing Technology Shock (solid line), Housing Demand Shock (starred line).

supply shocks explain each one-quarter of fluctuations in housing prices and housing investment. We show that only expectations of a future reduction in the supply of houses can generate boom-bust cycles in all aggregate quantities such as output, consumption, hours and investment. In contrast, expectations of a future increase in housing demand reduce business investment and therefore fail to generate comovement in aggregate variables. Fig. 6 (solid line) shows the response of key variables to expectations of a future decline in productivity in the housing sector, ie. a fall in Ah , that eventually does not realize. The perspective of a future slow down in the construction sector makes agents willing to increase their labor supply in order to mitigate the future negative effect of the shock on their choices. News of negative housing supply shocks generate expectations of a future increase in house prices. To take advantage of lower current prices, Borrowers increase their current demand for housing. Due to the collateral value of the house, Borrowers can expand their access to mortgage financing and increase consumption expenditure generating the extra amplification through the financial accelerator mechanism. Due to adjustment costs in capital, firms start adjusting the stock of capital already at the time of news.24 As a result, business investment slightly decreases on impact. Despite this, GDP rises due to the increase

in housing investment and consumption. A four-period anticipated decline in productivity in the construction sector generates a boom in housing prices, housing investment, consumption, GDP, hours and indebtedness. Still, current business investment slightly falls. At time t = 4, when agents realize that they were wrong about their beliefs, a sudden drop on all variables occurs, generating the hump-shape behavior in macroeconomic and housing variables usually observed during boom-busy phases. Fig. 6 (starred line) shows the response of the model economy to expectations of a future increase in housing demand due to a housing preference shock, an increase in j. For simplicity, also in this case we only report the response to news shocks that do not realize. Expectations of future housing demand shocks cannot generate the right co-movement between business and housing investment. The intuition is as follows. Anticipating a future increase in housing prices, Borrowers raise their current demand for houses and thus indebtedness and consumption. At the same time, due to an expected shift in preference for housing relative to consumption, firms in the housing sector start adjusting their stock of capital at the time of the signal. As a result, business investment falls. Despite the decline in business investment, GDP rises. Because of the reduction in business investment during the boom phase, news about a future increase in housing demand fails to generate boom-bust dynamics consistent with the empirical pattern. In the data business investment starts increasing on average six periods

24 The stock of capital (not shown in the graph) used as input of production in the consumption sector increases while it decreases in the housing sector.

307

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

and recessions.28 Qualitatively similar is the transmission of news on the central bank's inflation target. The anticipation of a higher target for inflation means higher long-run expected inflation. Firms that can change prices adjust them upwards already in the current period. Thus, expectations of higher future inflation increase inflation already in the current period. Expectations of a future reduction of the ex post real interest rate coupled with a current reduction in the nominal interest rate induce an increase in household indebtedness, higher consumption and higher housing spending. Housing prices and housing investment increase. Due to adjustment costs to capital, firms start adjusting the stock of capital already at the time of the signal. Real wages and hours worked rise. The economy experiences a macroeconomic boom. However, compared to the case of expectations of future expansionary monetary policy shocks, expectations of a temporary upward shift in the inflation target generate a less sizable boom but a more pronounced bust. In contrast, inflationary news shocks lead to a more sizable boom but a milder bust. See Fig. 7 (dashed line).

before the peak in housing prices. Expectations related to future housing demand make business investment decline throughout the boom phase. The behavior of business investment is independent of the time horizon of the expected increase in housing demand. Anticipated increase in housing demand at longer time horizons only postpones the occurrence of the peak. The decline in business investment is also robust to different parametrization of key model's parameters.25 5.2.2. News on monetary policy We also study the transmission of expectations of future monetary policy developments in driving business cycle fluctuations in the housing market. We document that expectations of a reduction of the policy rate or of an increase in the central bank's inflation target generate macroeconomic booms that turn into busts if agents' expectations are not realized ex post. Fig. 7 (starred line) reports the model's response to a standard unanticipated shock. An unexpected monetary policy loosening, ie. a negative realization of ϵR, t , leads to a decline in the interest rate that induces agents to increase their current expenditures. Aggregate demand rises. Borrowers significantly increase their level of indebtedness and housing investment. Housing prices rise and the subsequent collateral effect induces a sizable increase in Borrowers' consumption. In general, aggregate variables increase without displaying humpshaped dynamics. Anticipated monetary policy shocks can instead generate macroeconomic boom-bust dynamics. Fig. 7 (solid line) reports the effect of an expected four-period ahead one-period reduction in the policy rate of 0.1 percentage points, namely an anticipated shock to uR, t that does not realize. The intuition is as follows. Expectations of a future decrease in the policy rate generate expectations of a decline in the future real interest rate. Thus, borrowers anticipate this effect and increase their current consumption. Demand pressures rise current inflation. The ex post real rate declines reducing the current debt service. In addition, the anticipation of expansionary monetary policy also creates expectations of higher future housing prices that further induce Borrowers to increase their current demand for housing and thus indebtedness. Due to limits to credit, Borrowers increase their labor supply in order to raise internal funds for housing investments. Savers face a reduction in their current and expected interest income. Thus, for this group of agents consumption increases by less, current housing investment declines and their labor supply increases significantly. Given the adjustment costs of capital, firms in the consumption sector start adjusting the stock of capital already at the time in which news about a future reduction in the policy rate spread. For the increase in investment to be coupled with an increase in hours, wages rise in both sectors. GDP increases already at the time of the signal.26 In the case of expectations that do not realize there is a dramatic drop in both quantities and prices. Aggregate variables fall below their initial level. It takes about ten quarters for GDP to go back to the initial level. Expectations of looser monetary policy that do not realize generate a macroeconomic boom-bust cycle followed by a recession.27 Thus, good communication on monetary policy is essential for reducing the occurrence of expectations-driven cycles

6. Expected credit conditions and housing booms The results presented above show that the increase in housing prices generated by changes in households' expectations is coupled with an endogenous increase in household indebtedness. An oftenheard explanation for the last housing boom is the easing in credit conditions. Among others, Liebowitz (2009) finds that the most important factor related to foreclosures in the United States is the extent of negative equity in a home, which is directly related to low down payments. Indeed, in the years before the housing boom, mortgage credit in the US became more easily available to new home buyers.29 Piazzesi and Schneider (2009) looking at survey data report that good credit condition was the main driver of the last boom in its initial phase. Survey data reported in the previous section also show that expectations of future tighter credit are an important reason for consumers to believe that it is a good time to buy a house. In order to build some intuition of the effect of this type of news shock, we first document the effects of a current unexpected increase in the Loan-to-Value ratio.30 The dotted line in Fig. 8 shows the effect of a one percentage point temporary increase in the access to credit, namely a current increase in m. The increase in the current Loan-to-Value ratio allows borrowers to fund a larger share of consumption and housing with debt. This leads to a rise in aggregate consumption, investment and GDP and house prices. However, the model's responses do not display the hump-shaped dynamics that typically emerge in boom-bust cycles. The shock leads to an initial increase in house prices, investment, consumption and GDP and a subsequent, slow monotone decline toward the initial level. Second, we analyze the effect of favorable conditions in the credit market that are expected to be reversed in the near future. Fig. 8 (solid line) shows the effects of a one-percent current increase in m coupled with expectations of future restrictions in the access to credit, namely with expectations that m will return to its original value after four periods. For simplicity we analyze only the case in which expectations of a reversal in credit conditions are matched. Unmatched expectations would only generate a more sizable bust.

25 Expectations of a change in housing demand at different time horizons only postpone the occurrence of the peak. The behavior of business investment is independent of the time horizon of the expected increase in housing demand. The decline in business investment is also robust to different parametrization of key model's parameters. 26 We also consider the case of an expected persistent reduction in the policy rate. For this experiment we set the persistence of the shock uR, t equal to 0.65 in order to capture the situation where agents expect the policy rate to remain low for several periods. The effect on housing prices and on all other aggregate variables is stronger and the initial boom and the subsequent recession are more pronounced relative to the case where the expected reduction in the policy rate is only for one period. 27 In the case of an anticipated shock that realizes (not shown), aggregate variables would boom and then slowly decline. The peak response in output would correspond to the time in which expectations realize.

28 If agents expect the policy rate to remain low for several periods the effect on housing prices and on all other aggregate variables is stronger and the initial boom and the subsequent recession are more pronounced relative to the case where the expected reduction in the policy rate is only for one period. 29 Second-hand house market in the U.S. is relatively larger, but there are not long time series available to allow us to consider in our model. However, the new housing starts and the existing home sales follow a similar path, therefore we can abstract from the second hand house market, as the first market is a good proxy of the overall market. 30 To illustrate the effect of changes in the access to credit, we assume that m follows an AR(1) process with persistence equal to 0.994, as estimated by Iacoviello and Neri (2010). See Iacoviello and Neri (2010), Appendix D.

308

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

Fig. 7. News on monetary policy: unexpected monetary policy shock (starred line), unrealized monetary policy news (solid line), unrealized central bank's inflation target shock (dashed line).

mortgages increased from 70 percentage points in March 2004 to 80 percentage points in December 2007. In fact mortgages with LTV ratios of 0.7 or less went from 25 to 19 percent of the total while mortgages with LTV ratios above 0.9 increased from 15 to 30 percent of the total over the period 2004–2007.31 Duca et al. (2010) argue that the distinctive feature of the recent US mortgage lending and housing bubble was an unsustainable weakening of credit standards. By presenting some cross-country evidence they show that financial innovation further amplified the consumption effect of the bubble by altering the collateral role of housing.32 In the aftermath of the recent financial crisis, several countries have considered reductions in LTV ratios for mortgages to avoid the recurrence of house price and household debt cycles and mitigate potential vulnerabilities in the financial system. The U.S. Federal Housing Finance Agency is considering to increase down payment

Relative to the previous case, the impact on most variables is more sizable. Lower expected access to credit in the future induces Borrowers to increase their current demand for loans and housing more relative to the case analyzed above. As a result, the increase in housing prices and housing investment is more pronounced. Borrowers substitute consumption for housing and supply more labor in order to raise internal funds and take advantage of temporarily better access to credit. In contrast, Savers' consumption increases due to future lower real interest rates and habit persistence in consumption. Aggregate consumption increases as well as GDP and hours worked. Thus, in the absence of expectations of future reversal in credit conditions, the model's responses would not display the hump-shaped dynamics that typically emerge in boom-bust cycles. In contrast, a current easing of credit conditions that are expected to be reversed in the near future generates more pronounced and hump-shaped dynamics. 7. Boom-bust cycles and the LTV ratio

31 The Prudential Real Estate Investors (2009) reports that the estimated LTV ratios of new mortgages increased by 6 percentage points in the Euro area and 26 percentage points in the United Kingdom over the period 2004–2007. 32 Although LTV ratios for new mortgages vary significantly across OECD countries, they have risen until 2007. According to the ECB report on “Housing Finance in the Euro Area,” the typical LTV ratio for a first-time house buyer was around 80 percentage points in the Euro area in 2007, ranging between 63 and 101 percentage points. Similar ratios have been reported by Calza et al. (2013) and Cardarelli et al. (2009).

The previous section showed that expectations of future tightening in credit conditions can per se drive housing booms. Here we investigate the role of the LTV ratio for the transmission of expectations-driven cycles in our model generated by news on shocks originated in other sectors of the economy. The average LTV ratio for U.S. conventional single-family fixed-rate 309

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

Fig. 8. Unexpected increase in LTV (dotted line), temporary increase in the LTV ratio combined with expected future credit tightening (solid line).

prices and investment. These results are robust to considering unanticipated shocks only, news shocks only, or by reducing the standard deviation of news shocks to half the estimated standard deviation of contemporaneous shocks. However, it is worth mentioning that adding news shocks to the model dampens the reduction in the volatility of consumption, GDP and household debt and further increases the volatility of house prices and investment.34 A few papers have investigated the role of collateral requirements for the transmission of unanticipated shocks and macroeconomic volatility. Campbell and Hercowitz (2004) show that lower LTV increase the amplification of productivity shocks and thus imply higher volatility of output, consumption, and hours worked. According to their findings, the U.S. mortgage market liberalization of the early 1990s, proxied by an increase in the LTV ratio, played a role in explaining the great moderation. In contrast, Calza et al. (2013) show that the transmission of monetary policy shocks to consumption, investment and house prices is dampened by lower LTV ratio.35 Using a model with a richer stochastic structure that includes several types of

requirements for all mortgages. The Swedish Financial Supervisory Authority has recently introduced a maximum permitted LTV ratio of 0.85; tighter LTV ratios have also been implemented in Central and Eastern European economies, namely Latvia and Romania.33 Do lower LTV ratios reduce the severity of cycles in the housing market? In the following we explore the role of lower LTV ratios for macroeconomic volatility. We rely on the estimated model by Iacoviello and Neri (2010) augmented by a set of news shocks that generate the co-movement seen in the data during periods of housing booms. Table 6 reports the theoretical standard deviation of some key macroeconomic variables in our model for the benchmark value of the LTV ratio (m=0.85), a lower (m = 0.75) and a higher (m = 0.95) value. For this exercise we keep the LTV ratio constant at the specified value and set the standard deviation of news shocks equal to the estimated standard deviation of the contemporaneous shocks. A lower LTV ratio significantly reduces the volatility of households' debt, consumption and GDP. The effect is particularly strong for households' debt. In fact, compared to the benchmark LTV ratio of 0.85, its standard deviation is reduced by 30 per cent under a LTV ratio of 0.75. However, the stabilization effect of a lower LTV ratio on these variables is not accompanied by a reduction in the volatility of house

34 As documented in Appendix E, lower LTV ratios reduce the severity of both booms and busts in household debt, consumption and GDP. On the other hand, lower LTV ratios slightly amplify the cumulated fluctuations in housing prices, housing and business investment. 35 Walentin (2014a) quantifies the effects of higher LTV ratios in an estimated model of the Swedish economy and reports an increase of 8.3 and 24 percent in GDP and aggregate consumption, respectively, for an increase in m from 0.85 to 0.95.

33 For further details see the ECB report on “Housing Finance in Euro Area,” 2009. For a discussion of alternative policy tools, see the BIS report on “Macroprudential Policy Tools and Frameworks”, 2011.

310

Economic Modelling 60 (2017) 297–312

L. Lambertini et al.

Regarding the credit market, easier credit conditions that are expected to be reversed in the near future can generate boom-bust dynamics in line with the recent housing market cycle just before the financial crisis in 2007. Moreover, lower LTV ratios reduce the severity of booms and busts in GDP, consumption and households' debt. The model is solved under the case of binding constraints. An interesting extension for future research would be to explore the impact of news shocks under a nonlinear general equilibrium model with occasional binding collateral constraints in order to look at possible asymmetry generated by news shocks in explaining the boom-bust in the housing market.

Table 6 Macroeconomic volatility and LTV ratios.

Unanticipated shocks Household debt Consumption Business investment Housing investment GDP Housing prices Unanticipated +news shocks Household debt Consumption Business investment Housing investment GDP Housing prices

m = 0.95

m = 0.85

m = 0.75

% Difference (0.95−0.85)

% Difference (0.85−0.75)

0.2447

0.1218

0.0838

−50.238

−31.135

0.0182 0.0416

0.0162 0.0418

0.0156 0.0420

−10.945 0.427

−3.717 0.633

0.0837

0.0845

0.0850

0.897

0.665

0.0240 0.0218

0.0227 0.0218

0.0223 0.0219

−5.642 0.003

−1.569 0.537

0.2919

0.1614

0.1205

−44.695

−25.369

0.0314 0.0697

0.0289 0.0705

0.0278 0.0713

−8.193 1.159

−3.594 1.075

0.1362

0.1378

0.1388

1.149

0.754

0.0424 0.0311

0.0408 0.0315

0.0403 0.0318

−3.747 1.160

−1.167 1.040

Acknowledgments We are grateful to seminar participants at the Banco de España, the Banco de Portugal, the Bank of Finland, Universidade Nova de Lisboa, Catholic University of Louvain, LUISS, the University of Surrey, the meeting of the Society of Economic Dynamics, the 2010 meeting of the 16th International Conference on Computing in Economics and Finance, the European Economic Association 2010 Congress, the Fundação João Pinheiro and the Bank of International Settlements for useful feedbacks on this project. We also thank Pierpaolo Benigno, Isabel Correia, Daria Finocchiaro, Matteo Iacoviello, Peter Ireland, Nobuhiro Kiyotaki, Stefano Neri, Eva Ortega and Joao Sousa for valuable comments and suggestions. The last author acknowledges the Banco de Portugal for the 2009 visiting research grant. The opinions expressed in this article are the sole responsibility of the authors and do not necessarily reflect the position of the Banco de Portugal or the Eurosystem.

Standard deviations for HP-filtered series (λ=1600).

unanticipated and news shocks, we find that overall, lower LTV ratios sizably reduce the volatility of household debt, and also dampen variations in consumption and GDP, however, without mitigating fluctuations in house prices and investment. According to those results, LTV should be a key policy tool to dampen volatility in financial variables. Our results are in line with previous literature on the implementation of macroprudential policies. Angelini et al. (2010) show that, compared to capital-requirement rules, countercyclical LTV rules are more effective in reducing the variability of the debt-to-GDP ratio in response to productivity shocks. Christensen and Meh (2011) find that changes in the maximum LTV ratio are a more effective tool to mitigate boom–bust cycles in house prices than monetary policy since they directly target the main source of volatility. Mendicino (2012) draws implications on the use of the countercyclical LTV ratios as an alternative macro-prudential tool to monitor leverage, mitigate financial shocks and dampen the magnitude of the financial cycle.

Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.econmod.2016.10.004. References Adam, K., Kuang, P., Marcet, A., 2012. House Price Booms and the Current Account, NBER Macroeconomics Annual. University of Chicago Press, vol. 26, no. 1, pp. 77– 122. Adams, Z., Füss, R., 2010. Macroeconomic determinants of international housing markets. J. Hous. Econ. 19 (1), 38–50. Ahmad, N., Xiao, W., 2007. Prospect theory and the welfare cost of business cycles. Southwest. J. Econ. 10, 1. Angelini, P., Neri, S., Panetta, F., 2010. Macroeconomic stabilization policies: grafting macro-prudential tools in a macroeconomic framework, Manuscript. Barsky, R.B., House, C., Kimball, M., 2007. Sticky-price models and durable goods. Am. Econ. Rev., 984–998. Beaudry, P., Portier, F., 2004. An exploration into Pigou's theory of cycles. J. Monet. Econ. 51, 1183–1216. Beaudry, P., Portier, F., 2006. Stock prices, news, and economic fluctuations. Am. Econ. Rev. 96 (4), 1293–1307. Beaudry, P., Portier, F., 2007. When can changes in expectations cause business cycle fluctuations in neo-classical settings? J. Econ. Theory 135 (1), 458–477. Bjørnland, H.C, Jacobsen, D.H., 2010. The role of house prices in the monetary policy transmission mechanism in small open economies. J. Financ. Stab. 6 (4), 218–229. Burnside, C., Eichenbaum, M., Rebelo. S., 2012. Understanding Booms and Busts in Housing Markets, Meeting Papers – Society for Economic Dynamics, No. 114. Calza, A., Monacelli, T., Stracca, L., 2013. Housing finance and monetary policy. J. Eur. Econ. Assoc. 11 (s1), 101–122. Campbell, J.R., Hercowitz, Z., 2004. The Role of Collateralized Household Debt in Macroeconomic Stabilization. NBER Working Paper 11330. Cardarelli, R., Monacelli, T., Rebucci, A., Sala, L., 2009. Housing finance, housing shocks and the business cycle: evidence from OECD countries, Mimeo Christensen, C., Corrigan, P., Mendicino, C., Nishiyama, S.I., 2016. Housing collateral, residential investment and the Canadian business cycle. Can. J. Econ. 49, 1. Christensen, I., Meh, C.A., 2011. Countercyclical loan-to-value ratios and monetary policy, Manuscript Christiano, L., Ilut, C., Motto, R., Rostagno, M., 2010. Monetary Policy and Stock Market Boom-Bust Cycles. National Bureau of Economic Research, No. w16402. Davis, M., Heathcote, J., 2005. Housing and the business cycle. Int. Econ. Rev. 46, 751–784. Den Haan, W.J., Kaltenbrunner, G., 2009. Anticipated growth and business cycles in matching models. J. Monet. Econ. 56 (3), 309–327.

8. Conclusions We explore the transmission mechanism of news on business conditions as a source of optimism about the housing market and rising house prices. In our model expectations on a variety of shocks can generate housing-market booms in accordance with the empirical findings. In particular, housing booms are generated by news that put upward pressure on house prices, such as positive news on productivity in the non-durable sector, negative news on productivity in the housing sector, or news about a reduction in the policy rate and inflationary pressures; busts follow if the news are not realized ex post. News shocks definitely amplify the business cycle and re-produced the humpshaped response on housing and macroeconomic variables. A housing boom emerges when agents expect a future increase in housing prices, which fuels current housing demand and lifts housing prices and debt immediately. For example, an anticipated reduction in the policy rate generates expectations of lower borrowing costs, thereby raising immediately housing demand, mortgage debt and house prices. However, only expectations of monetary policy and inflationary shocks that are not fulfilled can generate macroeconomic recessions. 311

Economic Modelling 60 (2017) 297–312

L. Lambertini et al. Duca, J.V., Muelbauer, J., Murphy, A., 2010. Housing markets and the financial crisis of 2007–2009: lessons for the future. J. Financ. Stab. 6 (4), 203–217. Finocchiaro, D., Queijo Heideken, V., 2013. Do central banks react to house prices? J. Money Credit Bank. 45 (8), 1659–1683. Floden, M., 2007. Vintage Capital and Expectations Driven Business Cycles. CEPR Discussion Paper 6113. Foote, C., Gerardi, K., Willen, P., 2012. Why did so many people make so many ex post bad decisions? The causes of the foreclosure crisis. National Bureau of Economic Research, No. w18082. Gomes, S., Iskrev, N., Mendicino, C., 2014. Monetary policy shocks: we got news, Mimeo Greenwood, J., Hercowitz, Z., G. W., Huffman, 1988. Investment, capacity utilization, and the real business cycle. Am. Econ. Rev., 402–417. Huang, M., 2014. Bubble-like housing boom-bust cycles: evidence from the predictive power of households' expectations. Q. Rev. Econ. Finance 54 (1), 2–16. Kanik, B., Xiao, W., 2014. News, Housing Boom-Bust Cycles, and Monetary Policy. State University of New York (October 2009) Iacoviello, M., 2005. House prices, borrowing constraints, and monetary policy in the business cycle. Am. Econ. Rev. 95 (3), 739–764. Iacoviello, M., Minetti, R., 2006. International business cycles with domestic and foreign lenders. J. Monet. Econ. 53 (8), 2267–2282. Iacoviello, M., Neri, S., 2010. Housing market spillovers: evidence from an estimated DSGE model. AEJ Macroecon. 2 (2), 125–164. Int'Veld, J., Raciborski, R., Ratto, M., Roeger, W., 2011. The recent boom-bust cycle: the relative contribution of capital flows, credit supply and asset bubbles. Eur. Econ. Rev. 55 (3), 386–406. Jaimovich, N., Rebelo, S., 2009. Can news about the future drive the business cycle? Am. Econ. Rev. 99 (4), 1097–1118. Khan, H., Tsoukalas, J., 2012. The quantitative importance of news shocks in estimated DSGE models. J. Money Credit Bank. 44 (8), 1535–1561. Kiyotaki, N., Moore, J., 1997. Credit cycles. J. Polit. Econ. 105 (2), 211–248. Kobayashi, K., Nakajima, T., Inaba, M., 2007. Collateral Constraint and News-driven Cycles. Discussion Paper 07013, Research Institute of Economy, Trade and Industry (RIETI). Kuang, P., 2014. A model of housing and credit cycles with imperfect market knowledge. Eur. Econ. Rev. 70, 419–437. Lambertini, L., Mendicino, C., Punzi, M.T., 2013. Expectation-driven cycles in the

housing market: evidence from survey data. J. Financ. Stab. 9 (4), 518–529. Lambertini, L., Mendicino, C., Punzi, M.T., 2013. Leaning against boom-bust cycles in credit and housing prices. J. Econ. Dyn. Control 37 (8), 1500–1522. Leamer, E., 2015. Housing really is the business cycle: what survives the lessons of 2008–09. J. Money Credit Bank. 47 (S1), 43–50. Mendicino, C., Punzi, M.T., 2014. House prices, capital inflows and macroprudential policy. J. Bank. Financ 49, 337–355. Liebowitz, S., 2009. Anatomy of a train wreck: causes of the mortgage meltdown. In: Housing America: Building Out of a Crisis, vol. 287, pp. 292–298 Liu, Z., Wang, P., Zha, T., 2013. Land-price dynamics and macroeconomic fluctuations. Econometrica 81, 1147–1184. Mendicino, C., 2012. Collateral Requirements: Macroeconomic Fluctuations and macroPrudential Policy. Banco de Portugal Working Papers 11-2012. Milani, F., Treadwell, J., 2012. The effects of monetary policy news and surprises. J. Money Credit Bank. 44 (8), 1667–1692. Nofsinger, J.R., 2012. Household behavior and boom/bust cycles. J. Financ. Stab. 8 (3), 161–173. Nutahara, K., 2010. Internal and external habits and news-driven business cycles. Econ. Lett. 107 (2), 300–303. Piazzesi, M., Schneider, M., 2009. Momentum traders in the housing market: survey evidence and a search model. Am. Econ. Rev. P & P 99 (2), 406–411. Prudential Real Estate Investors, 2009. Deleveraging Commercial Real Estate in Europe. Report. Schmitt-Grohe, S., Uribe, M., 2012. What's news in business cycles. Econometrica 80 (6), 2733–2764. Soo, C., 2013. Quantifying Animal Spirits: News Media and Sentiment in the Housing Market. The Stephen M. Ross School of Business at the University of Michigan Research Paper Series. Tomura, H., 2010. International capital flows and expectation-driven boom-bust cycles in the housing market. J. Econ. Dyn. Control 34 (10), 1993–2009. Walentin, K., 2014a. Expectation driven business cycles with limited enforcement. Econ. Lett. 124 (2), 300–303. Walentin, K., 2014b. Housing collateral and the monetary transmission mechanism. Scand. J. Econ. 116 (3), 635–668. Walker, C., 2014. Housing booms and media coverage. Appl. Econ. 46 (32), 3954–3967.

312