Are house prices responsible for unemployment persistence?∗ Ekkehard Ernst† Research Department International Labor Organisation
Faten Saliba Department of Economics University of Manchester
Autumn 2016
Abstract
The paper analyses the ambiguous role of house prices and housing investment for unemployment dynamics. Whereas traditional models see an increase in house prices as a dynamic multiplier that contributes positively to business cycle swings, the paper considers additional transmission mechanisms via the competitiveness channel (wages) and productivity. As house prices rise, wages tend to follow to make up for the loss in real disposable income, which limits employment creation. In addition, with rising house prices, the relative size of the construction sector a low-productivity industry tends to increase, lowering aggregate productivity growth, further dampening competitiveness. The paper estimates a stylised dynamic general equilibrium model with unemployment ows. Introducing dierent transmission mechanisms through which the housing market in uences labour and macroeconomic dynamics, the size and direction of the housing market channel is being analysed. The estimation results show that housing shocks can have long-lasting negative eects on employment even though a housing boom can generate a short-lived stimulus on growth and employment. The paper also oers some policy advice simulating housing shocks under dierent types of structural reforms and macro-prudential regulation. Keywords: Housing market, house prices, unemployment, competitiveness, asset price booms JEL-Codes: J64, J23
∗ The authors would like to thank Michael Kumhof, Immo Schott as well as participants at the 7th ReCapNet Conference, a Joint IAB-University of Nuremberg labour market seminar, the 2013 Project LINK meeting and a monthly IMF Labour Market Seminar for very insightful comments. All remaining errors are ours.
† Corresponding author, e-mail:
[email protected]
1
1 Introduction House price developments have long been neglected by standard economic models. Housing capital has been seen as simply a part of the overall accumulated capital stock, without any separate role to be played by the housing wealth in the overall business cycle dynamics. With the rapid rise in house prices in the early 2000s, especially in the United States but also in some European countries, researchers started to pay closer attention to any business cycle eects this might have.
1
With
housing representing the biggest part of household wealth, changes in the price of a house was seen as a potentially powerful source of business cycle uctuations. And indeed, both theoretical and empirical studies demonstrated the large impact that housing wealth can have on aggregate consumption as households feel richer and increase their consumption as house prices rise. With the onset of the global nancial crisis that was triggered by a sharp correction in housing prices, the interest in the linkages between housing dynamics and macroeconomic performance increased
2
signi cantly.
House prices often have a life of their own.
3
High transaction costs and taxes as well as the
speci c structure of the mortgage market can lead to a drawn-out collapse of the housing market when the housing price cycle turns. The adjustment costs of the housing stock to variations in demand, if sudden and large, can take a substantial amount of time. In Japan, for instance, housing markets only gave signs of a revival two decades after they had collapsed in the early 1990s. Similar phenomena can be observed several years after the start of the great nancial crisis in 2008/09. In Spain, for instance, house prices were still 20 per cent higher in 2013 than before the start of the house price boom in the early 2000s despite the fact that the country went through the worst and longest-lasting recession in its recent history. In France, house prices even stabilised at historically high levels against the background of an essentially stagnating economy. In other countries, house prices continue to rise, most prominently in Germany where real house prices are now close to their historical peak in the 1980s (see gure 1). Such large swings in housing prices are now increasingly seen as a potentially destabilizing force in the economy. In particular, some observers have argued that high house prices can at least partly explain the large and persistent increase in unemployment in these countries following the burst of the housing bubble. Speci cally, high house prices translate into stronger wage increases as living costs rise. Also, higher house prices especially when they aect commercial property price increase capital costs for rms, reducing investment and job creation (Askenazy, 2013). Moreover, in the aftermath of the crisis and with dierent regions aected to various degrees, sluggish geographical mobility was seen as one of the factors behind the persistence in US unemployment rates (see, for instance, Estevão and Tsounta, 2011).
1 See, for instance, Iacoviello (2005); Iacoviello and Neri (2007) 2 See, for instance, Iacoviello and Pavan (2013); Jordà et al. (2016). 3 For a historical, global comparison of house prices, see Knoll et al.
(Forthcoming), for a comparison of house
price dynamics in Europe see Corradin and Fontana (2013), and for an analysis of recent trends in China see Fang et al. (2015).
2
140
Figure 1: Recent house price developments Brazil
Germany China
OECD
80
Real house prices (2010 = 100) 100
120
United States
France
60
United Kingdom
2000q1 Note:
2005q1
2010q1
2015q1
The chart shows recent real house price developments in selected
OECD and emerging economies. Real house prices are indexed at 100 in 2010. Source: OECD, House price database, October 2016
This is in contrast to earlier, positive views on the role of housing dynamics for economic growth. For a long time, house price increases have been welcomed as a positive driver for growth. As house prices rise, making private home owners feel richer, private consumption is expected to expand faster than economic activity thanks to the wealth eect (see Boone and Girouard, 2002; Iacoviello, 2011). Also, stronger house price increases allow for faster employment creation as they will lead to the expansion of the construction sector, an employment-intensive industry. Along these lines and prior to the crisis, many studies found that the housing and non-residential investment were positively comoving and strengthening the business expansion (Davis and Heathcote, 2005). This paper aims at disentangling the dierent transmission mechanisms through which the swings in housing prices aects the macro-economy and, speci cally, employment growth.
In
order to estimate the impact of the housing cycle on the labour market, a small dynamic general equilibrium model has been set up for a panel 14 OECD countries.
The model contains four
channels of how housing investment and house prices impact on labour market dynamics: (i) House price in ation leads to faster wage growth; (ii) housing investment pushes aggregate productivity down as the size of construction in the overall economy increases; (iii) housing investment leads to faster job creation and lower job destruction as construction is an employment-intensive industry; and (iv) house price increases strengthen aggregate demand via their eect on housing wealth. At the same time, the model also includes a channel for general asset price increases (both housing and share prices) to aect gross xed capital formation and long-term interest rates. All parameters of the model are estimated and an average eect of dierent shocks is evaluated across the countries included in the sample. A linear version of the model has been estimated using general methods of moments with full
3
information to get realistic (i.e.
not calibrated) parameters.
In order to prevent the estimated
parameters from being aected by the recent (global) house price cycle, the estimation sample has been limited to the period prior to 2007.
The baseline simulations show realistic reactions
with respect to productivity, housing investment and asset price shocks, in line with results in the (empirical) literature. Moreover, the model simulation shows that house market shocks, even though they might have short-term positive eects, lead to increases in unemployment over the medium term. Pure house price shocks are shown to increase unemployment almost instantaneously despite their (weak) eect on aggregate demand. Housing investment, on the other hand, has a short-lived- positive impact on the labour market, which is, however, followed by a long, drawnout period of joblessness and under-investment. In particular, the competitiveness eect on both higher wages and lower productivity growth is leading to a signi cant increase in unit labour costs, with particularly deleterious eects on the labour market. The paper also considers alternative policy responses to prevent such long, drawn-out labour market slumps. In particular, it compares ex-ante structural reforms that limit the pass-through of house price increases on wages with ex-post leaning-against-the-wind approaches to monetary policy (see, for instance, Kuttner and Shim, 2013).
The simulation experiments in this paper
demonstrate that the latter approach to the housing cycle is much more eective in addressing the overshooting in asset prices (not only house prices) than reducing wage rigidities through, for instance, structural reforms on labour markets. The paper is organised as follows:
The next section presents an overview of the literature.
Section 3 introduces the core labour market equations as well as the macroeconomic relations. Section 4 gives an overview of the data and methodology used. Section 5 discusses the estimation results whereas section 6 simulates the model and introduces policy simulations. A nal section concludes.
2 Housing markets and the macro-economy Increases in house prices have long been treated similarly to a general increase in asset prices. Based on considerations by Phelps (1998), Zoega (2009) argued that asset price dynamics indicate future expectations of economic performance and are hence negatively correlated with the unemployment rate. In his study he demonstrated that there is an empirical, negative medium-term correlation between unemployment and share prices. Focusing more speci cally on the housing market, Liu
et al. (2013) oer a structural analysis of the linkages between the housing and the labour market by adopting a DSGE model with credit and search frictions, relying on US macro time series data. In line with a wealth eect of housing prices on the macro-economy, their model suggests that there exists a negative correlation between the land price and the unemployment rate throughout the business cycle. In addition, they claimed that a negative shock to the land price will increase the level of unemployment and will also decrease the level of consumption, investment, total hours, and vacancies. Their results revealed that by decreasing the land price by 10 per cent, unemployment rate increases by 0.34 per cent. This structural analysis is crucial for understanding business cycles and policy analysis. The crisis demonstrated, however, that the pro-cyclical nature of the impact of asset and house prices on the real economy might by asymmetric (Lee and Chen, 2015), not least
4
because of a tendency for over-accumulation that frequently appears to accompany investments on asset markets (Hoon, 2010). One of the rst to look more closely into the speci c transmission mechanisms through which the housing market aects the rest of the economy, Iacoviello and Neri (2010) examined two important factors of housing. From the supply side, they examined sectoral heterogeneity which permits to catch the distinct movement and cyclical characteristics of housing prices and investment relative to other prices and to other forms of demand. From the demand side, they focused on the collateral eects of housing prices on borrowing which enables spillovers from the housing market onto consumer spending. More speci cally, the authors analyzed the causes and consequences of variations in the US housing market. During the last forty years, the housing market has witnessed only a slow technological progress, which justi es the existing upward trend in real housing prices. Housing demand and technology shocks each describe one-quarter of the instability of housing prices and investment. Monetary factors had an in uential role; however they described less than 20 per cent of the volatility of housing investment and prices. Last, they show in their study that over time spillovers from the housing market on consumption rather than on investment cannot be neglected. Branch
et al. (2016) developed a searching-matching model integrating the labour market, the
housing market and frictional goods market together. They examined the long-term in uence of household nance on both the labour market and housing market. Their general equilibrium model comprises several dimensions: First, the labour market consists of both a non-housing sector and a construction sector. Second, in the frictional goods market, household consumption is nanced with collateralized loans. Third, in the housing market, households are capable of renting housing services as well as buying and selling homes. The results of this model showed the existence of multiple steady states across which unemployment and housing prices are negatively correlated. In addition, this result has been used to analyze how variations in lending standards could have an in uence on the whole economy. Unaordable housing has been shown to hamper (local) employment growth, looking at dierent municipalities in California (Chakrabarti and Zhang, 2015). Measuring housing aordability as a share of households' disposable income, the authors demonstrate how high house price increases in California have hampered local employment growth. The paper does not distinguish, however, whether these eects are driven by fast rising wages (demand factors) or slower labour supply growth (supply factors).
An alternative strategy is used by Askenazy (2013) who considers the
eect of house prices on rms' competitiveness directly. He argued that a rise in the prices of lands, buildings and structures belonging to private companies could impede a company's competitiveness. First, rising commercial property prices are making investment for rms too costly. Second, by increasing the price of non- nancial assets, which at the same time increases the value of equity, rms will have to raise their dividends in order to maintain their owners' direct payments. This will decrease the incentive of rms to oer new vacancies. In his study, (Askenazy, 2013) relies on French national accounts and argues that the nominal value of buildings and lands has surpassed its historical value. As a result, non- nancial corporations in France have to pay a substantial extra cost for their investments and in turn have to allocate signi cant additional dividends. The impact house price cycles have on educational choices further compounds these competitiveness eects:
5
Indeed, looking over the entire episode of the recent house price boom and bust in the United States, Popov and Laeven (2016) demonstrate that the rapid increase in wages during the boom period raise opportunity costs for education (and hence lower incentives for college education) whereas the ensuing bust in house prices further depresses educational investment as falling housing wealth stiens budget constraints to invest in long-term studies (such as a college degree). House price dynamics can also aect the matching e ciency on the labour market, thereby lowering out ows out of unemployment into employment and hence overall employment growth. Large shares of (public) home ownership and rent-controlled housing have long been considered to be a barrier for geographical mobility and labour market adjustment to shocks (see, for instance, Hughes and McCormick, 1987, for the UK). In the aftermath of the crisis, this mechanism has been looked at anew in the United States, where Blanch ower and Oswald (2013) show that across US states a rise in the rate of house ownership leads to a signi cant increase in the unemployment rate in that state. Their study argued that a doubling of any particular US state's house ownership will lead to an increase in the unemployment rate by more than double its current value over the long run. This is explained by arguing that the increase in home ownership will decrease labour mobility and will hinder the opening of new businesses as house owners will have di culties to move to another location or open another business when the value of their home has declined substantially. Similarly, Farber (2012) shows that the failure to sell their house due to the depressed state of the housing market during the Great Recession prevented many unemployed to move to a dierent city or state in order to take up an alternative job opportunity after having lost the current one. These eects seem to be particularly relevant during periods of high unemployment whereas in an upswing both labour and housing markets are su ciently liquid for home ownership to have little eects on unemployment spells (Head and Lloyd-Ellis, 2012). On the other hand, lowering commuting costs can signi cantly decrease the impact of housing frictions on unemployment dynamics (Rupert and Wasmer, 2012). Overall, therefore, it seems that housing frictions and geographical mobility seem to play only a minor role in explaining unemployment dynamics but a potentially signi cant one during periods of downturns, con rming the asymmetric nature of the interaction between housing and labour markets, as discussed above. Finally, understanding the impact of housing dynamics on the real economy also requires to look at the overall credit cycle. As highlighted by Jordà
et al. (2016), recent increases in (global)
house prices have been accompanied by a signi cant extension of mortgage nance, substantially above historical standards.
Rünstler and Vlekke (2016) show that this expansion in credit has
fueled the house price increase in major advanced economies.
This rapid growth in credit was
mostly triggered by lower lending standards, more than by an endogenous weakening of borrowing constraints, at least in the United States (Justiniano
et al., 2015).4
Not only did this rapid expan-
sion of housing credit fuel the previous boom, it also alters the transmission channel for monetary policy, potentially increasing the eectiveness of monetary policy, at least in those countries where mortgage markets are exible, where mortgage equity can easily be released and where there is a large share of variable interest rates among mortgage contracts (Calza
et al., 2013).
Under these
conditions, leaning-against-the-wind monetary policies can be eective approaches to deal with
4 Several
authors have argued that this weakening in lending standards was used as a substitute social policy
to expand home ownership among lower-income households as a means of addressing rising inequality (see Rajan, 2010; Kumhof et al., 2015).
6
house and asset price swings to prevent longer-term damage in the real economy (Kuttner and Shim, 2013), a point we will also stress in this paper.
3 A housing market model 3.1
Model overview
In order to analyse the eect of housing markets on labour market dynamics, a linear DSGE model is augmented with a housing channel and an explicit formulation of the labour market. The model consists of three parts: The rst part is an empirical formulation of a standard search-andmatching model of the labour market, based on the earlier contribution by Carlsson
et al. (2006).
The approach allows a full and separate account of unemployment in- and out ows, rst used in a macroeconomic set-up by Ernst and Rani (2011). Dierent to a standard labour market demand equation, this one accounts explicitly for sluggish employment adjustment through opening of vacancies, hiring and ring dynamics. The labour market is embedded in a standard New Keynesian Phillips curve (NKPC) augmented with a wage bargaining curve resulting from the Nash bargaining process between rms and workers. Similar to the set-up used by Flaschel
et al. (1997) and Erceg et al. (2000),
such a
double Phillips curve allows for explicit wage-price dynamics. In our set-up, however, the linear nature of the model does not allow for complex dynamics as the ones studied by Flaschel
et al.
(1997). The housing market is modeled separately through a housing investment and house price equation. Housing aects both the macro-economy and the labour market through four channels: First, a standard wealth eect whereby higher house prices increase households wealth and the propensity to consume; second, through an aggregate demand eect whereby housing investment leads to job creation and lowers job destruction; third, through a wage eect whereby higher house prices pushes wage earners to demand for higher salaries in order to compensate for increases in living costs; and forth, through a productivity eect, assuming that the housing capital stock is less productive than the non-residential capital stock (see, e.g. Askenazy, 2013, and Iacoviello, 2010). The empirically estimated model is then shocked to analyse the eect of dierent policy scenarios.
In particular, short- and long-term eects of house price dynamics are being analysed
under the assumption that (a) labour market policies lower the pass-through of house prices on wage dynamics (through reduced bargaining power) and (b) that macro-prudential policies adjust
5
short-term interest rate in a leaning-against-the-wind fashion.
3.2
Labour market dynamics
Testable relations between macroeconomic variables such as housing and unemployment ows can be set up starting from a labour ow accounting framework. In particular, in the following, we concentrate on labour market ows related to ows from employment to unemployment (INt ) and
5 See
section 6 for the detailed simulation methodology and results.
7
from unemployment into employment (OU Tt ). These ows can be linked to (absolute) changes in the number of unemployed as follows:
4Ut = 4Lt − 4ETt = INt − OU Tt
(1)
i.e. the level of unemployment increases with the increase in the labour force, with the rise in (total) employment. into the unemployment pool,
INt ,
ETt .
LFt ,
and decreases
Alternatively, unemployment increases when in ows
exceed out ows,
OU Tt .
In order to be operational for our purposes, this ow equation needs to be further re ned, taking into account the job creation and destruction process that aects the total amount of jobs available:
4ETt = JobCreationt − JobDestructiont
(2)
i.e. changes in the employment level result from the dierence between created vs. destroyed jobs. Following standard labour market matching models (see Pissarides, 2000, for a classical reference in this area, and Carlsson
et al., 2006), the extensive margin of labour demand6
a mix of the following factors:
is determined by
7
L JobCreationt = α1 + β11 ETt−1 + β12 wt + β13 ADt + β14 Ut + β15 Vt + β16 rt−1 where
Vt :
ETt−1 :
past employment,
L un lled vacancies, rt−1
:
wt :
real wages,
ADt :
aggregate demand,
(lagged) real long-term interest rates and
α1 :
Ut :
(3)
unemployment,
a constant.
Similarly, job destruction will be aected by the rate of technological progress, the real interest rate (through the discounted future bene ts of an ongoing relationship), import competition, wages and aggregate demand:
JobDestructiont = α2 + β21 T F Pt + β22 rtL + β23 T oTt + β24 IM Pt + β25 wt + β26 ADt rtL :
(4)
where
T F Pt :
T oTt :
the real eective exchange rate (i.e. the terms of trade) to re ect import competition and
an indicator for total factor productivity,
the real (long-term) interest rate and
its eect on job destruction. Finally, unemployment dynamics are also aected by changes in the labour force. We re ect standard theories about the determinants of labour supply by considering the following equation for changes in labour supply (see, for instance, Burniaux
et al., 2003):
4Lt = α3 + β31 4Lt−1 + β32 4ut−1 + β33 T axt where
β32
(5)
represents the discouraged worker eect which depresses labour force growth (with
an expected negative sign).
Wages in the above equations will be determined through a wage
bargaining process to be discussed below. The ve equations (1), (2), (3)-(5) form the basis of our labour market ow model.
6 In this set-up, we abstract from changes in working hours. 7 Here as in the following, equations a written in their linearized, notational convenience.
8
Due to
empirical form. Error terms are dropped for
the lack of internationally comparable data on job creation and destruction rates, however, the model needs to be rewritten to match it with our database.
This can be done by bringing in
accordance job creation rates and unemployment out ows on the one hand, and job destruction and unemployment in ows on the other. This requires that the determinants of labour supply as speci ed in equation (5) are plugged into the appropriate unemployment ow equation. Indeed, unemployment in ows and out ows do not match exactly job destruction and job creation.
8
Some
unemployment in ow happens from inactivity (see chart) when the economy recovers while some people who are losing their job might drop out immediately into inactivity if they do not qualify for any bene ts. Similarly, job creation can happen out of inactivity (for instance through selfemployment), while some people might ow out of unemployment at the end of their bene t period and into inactivity. As a consequence, using unemployment ows instead of job creation and destruction rates might overestimate employment dynamics due to the failure to take out ows back and forth from and to inactivity. It might also overestimate the variation of employment growth when the inactivity rate uctuates with the business cycle (as is suggested by the discouraged worker eect). In our speci cation, we consider that the discouraged worker eect will create additional unemployment out ows. On the other hand, increasing supply in the available workforce will show up as additional unemployment in ows. Tax-related changes in labour supply are considered to aect unemployment in ows, in particular.
Besides these adjustments to our speci cation, we
consider both unemployment in ows and out ows to follow dynamic adjustment processes, instead of estimating them in levels. This way, we cope with systematic under- or overestimation of ows over the cycle that are due to these linkages between unemployment and inactivity. In addition, by considering contemporaneous interactions between the two ow directions, we also take care of the possibility that we are overestimating the impact of unemployment ows on employment variation: Higher contemporaneous in ows will also increase out ows as part of it goes into inactivity.
Similarly, higher out ows might partly imply an increase in inactivity that will show
up in increased in ow rates.
We will therefore estimate the following two equations related to
unemployment dynamics:
where
OU Tt
=
αOU T + β1OU T · OU Tt−1 + β2OU T · XtJobCreation + β3OU T 4ut−1
(6)
INt
=
αIN + β1IN · INt−1 + β2IN · XtJobDestruction + β3IN 4Lt−1
(7)
XtJobCreation
and
XtJobDestruction
correspond to the dierent explanatory variables in
equations (3) and (4) respectively. This will form the base model for the following extensions of our labour ow model.
8 See,
for instance, Bassanini et al. (2010); Davis et al. (2012) for a discussion between labour and worker ows.
9
3.3
The macro-economy
3.3.1 Capital accumulation and nancial frictions Financial frictions are introduced in the capital formation process through two channels: First, we assume a nancial accelerator eect in the spirit of Bernanke
et al. (1999) and Phelps (1998).
Speci cally, we use the set-up suggested by Phelps (1998) whereby share price dynamics,
L gether with real long-term interest rates, rt , and labour productivity growth,
LPt ,
Ft ,
to-
aect (private
sector, non-residential) capital accumulation through their impact on nancing conditions for businesses. In addition, the dynamics of gross xed capital formation depends on (externalities from) public investment,
GovInvt ,
and housing investment,
ItHousing .
With the latter, we follow the
two-sector set-up originally suggested by Iacoviello and Neri (2010) :
Housing L ∆Kt = αK + β1K Ft−1 + β2K rt−1 + β3K LPt−1 + β4K GovInvt−1 + β5K It−1
(8)
Moreover, interest rate dynamics are following a yield curve arising from market segmentation and dierent risk premiums attached to dierent maturities. Short- and long-term interest rates then depend on variations in key macroeconomic variables (see, e.g., Wu, 2003; Dai and Philippon, 2005). The long-term interest rates is linked to the policy rate via the yield curve, creating a wedge between long- and short-term rates.
Short-term rates are determined by policy interventions;
long-term rates, in contrast, react (sluggishly) to nancing conditions (i.e. share prices) and the sovereign risk premium via net government lending,
N LGQt .
The latter refers to endogenous
risk premiums that arise from over-leveraging and has been shown to have a strong in uence on short-term dynamics and the reactions of the real economy to shocks (see 2010; Ernst
Ernst and Semmler,
et al., 2016a): L rtL = αrL + β1rL Ft + β2rL rt−1 + β3rL rtS + β4rL N LGQt
(9)
3.3.2 Housing investment, house prices and aggregate labour productivity We follow Iacoviello (2010); Iacoviello and Neri (2010) in the way the housing market is being modeled. Housing investment (as a share of GDP),
ItHousing ,
depends on demographics, nancing
conditions, and the business cycle as measured by the output gap:
Housing L ItHousing = αIH + β1IH It−1 + β2IH rt−1 + β3IH ∆LFt−1 + β4IH Gapt−1 As a proxy for demographic dynamics we will use labour force growth,
∆LFt ,
(10)
assuming that a
faster growth in the labour force (in comparison to a purely demographic change) is needed to stimulate demand for housing. The variation in house prices,
PtHousing ,
in turn, is supposed to depend on past price increases
and housing investment:
Housing Housing ∆PtHousing = αP H + β1P H ∆Pt−1 + β2P H It−1
10
(11)
Construction is assumed to suer from low(er) productivity growth in comparison to other sectors. In particular, residential investment is assumed to be less productive than non-residential investment.
This creates a composition eect when the share of the residential capital-to-total
capital increases, lowering the productivity growth rate of the total economy. Labour productivity growth, therefore, is assumed to accelerate following innovations in total factor productivity and but expected to be dragged down by housing investment:
∆P rodt = αP rod + β1P rod ∆P rodt−1 + β2P rod ∆T F Pt + β3P rod · ItHousing
(12)
3.3.3 Wage bargaining, price in ation and aggregate demand The rest of the macro-economic set-up follows Ernst and Rani (2011). Speci cally, (sluggish) wage growth,
∆W agest , results from a bargaining set-up that shares the bene ts of productivity growth,
taking into account expected price in ation as well as - speci cally - house price developments:
Housing ∆W agest = αw + β1w W agest−1 + β2w ∆P rodt−1 + β3w Eπt+1 + β4w ∆Pt−1 + β5w ∆Ut
Price in ation, exchange rate, looking
πt ,
9
(13)
is in uenced by imported in ation through changes in the real eective
REERt ,
but otherwise follows a hybrid NKPC with both backward and forward-
10 elements:
πt = απ + β1π πt−1 + β2π Eπt+1 + β3π REERt−1 + β4π Gapt−1
(14)
Finally, aggregate demand is measured by the dynamics of the output gap that follows labour market dynamics, private investment and government spending. Moreover, includes a wealth eect from house price appreciations:
Housing Gapt = αGap +β1Gap wt +β2Gap OU Tt−1 +β3Gap INt−1 +β4Gap ∆Kt +β5Gap GovConst−1 +β6Gap Pt−1 (15)
3.3.4 Monetary and scal policy Following speci cations suggested by,
inter alia, Judd and Rudebusch (1998) and Woodford (2001),
monetary policy is introduced via the Taylor rule, assuming interest rate smoothing and linking the (nominal) short-term interest rate,
iSt ,
to price and output gap dynamics:
iSt = αrS + β1rS · iSt−1 + β2rS · Gapt + β3rS · πt−1
11 (16)
Net government lending in equation (9) and government consumption in equation (15) is endogenised, considering scal policy rules that link the state of the labour market to government
9 For a discussion of dierent theories of wage determination see Ernst et al. (2016b). 10 The hybrid NKPC is better be thought of as resulting from sticky information,
similar to the framework
developed by Mankiw and Reis (2002).
11 In
the estimation in section 5, we will make use of the Fisher equation to link the real to the nominal short-term
interest rate, i.e.
rtS = iS t − πt .
11
spending and revenues (see table 1):
Table 1: The government account Government consumption
GovConst = β1GCons GovConst−1 + β2GCons ∆ETt
Government investment
GovInvt = β1GInv GovInvt−1 + β2GInv ∆ETt−1
Tax revenues
T axt = β1T ax · OU Tt−1 − β2T ax INt−1
Net government lending
N LGQt = T axt − GovConst − GovInvt
Note that
N LGQt
is de ned such that
N LGQt < 0
represents a public sector de cit.
4 Data and methodology 4.1
Data
The paper brings together three databases: Unemployment in- and out ows, general macroeconomic data and house price information. The resulting database covers 14 OECD countries over most of the period between 1970 and 2007 on an annual basis (time coverage changes depending on the particular speci cation used). In order to eliminate the speci c eects of the global nancial crisis, no data post 2008 has been included. Unemployment ows are based on a methodology developed by Shimer (2012) and Elsby
et al.
(2013). The data is constructed on the basis of OECD information regarding unemployment stocks and unemployment duration at dierent duration lengths. In particular the data by Elsby
et al.
(2013) allow for a systematic cross-country analysis. In our case, we take advantage from the larger country coverage to use the increased number of degrees of freedom (within a panel-data context) in order to test for a larger variety of determinants of unemployment ows. Data is available in
12
the ILO Key Indicators of the Labour Market (KILM) repository.
House prices are taken from the OECD House price database that is updated monthly and available for 35 developed and emerging economies. The database has been complemented with information taken from the OECD Economic Outlook database and the OECD Main Economic Indicators. In particular, data regarding total employment and labour force developments, capital stock estimates and interest rates are taken from there. In addition, an indicator of real share price increases has been developed on the basis of OECD information, using the GDP de ator to de ate nominal share prices. Summary statistics of the key labour market and macroeconomic variables used in this paper can be found in table 6 in Appendix A.
4.2
Estimation methodology
The model set up by equations (6)-(16), together with the accounting equation (1) and the government accounts in table 1, constitutes a fully speci ed macroeconomic model with price- and
12 ILO
KILM: www.ilo.org\kilm
12
wage-Phillips curves, an Okun's law and an aggregate supply equation. In order to simulate its dynamic behaviour, we rst need to estimate the coe cient matrix
α β
= =
[α, β]
where
(αOU T , αIN , αK , αrL , αIH , αP H , αP rod , αw , απ , αGap , αrS )
T
β OU T , β IN , β K , β rL , β IH , β P H , β P rod , β w , β π , β Gap , β rS
and
T
.
Given the limited amount of observations for each country due to the annual nature of our database, we pool the data into a panel but remove country-speci c xed eects using de-meaning. Hence all our variables are transformed in the following way:
dXit = Xit − Xi· + X·· where a subscript dot indicates an average across countries or over time. This gives us a total of 247 observations for the estimated model of section 5. In order to estimate the coe cient matrix, we use a system-GMM estimator on the transformed variables, thus allowing us to control for endogeneity. Given the long time series and large country variations in our panel, we control for heteroscedasticity and auto-correlation, using Newey-West standard errors. The model is estimated as a simultaneous equation model with error terms:
T
ε = (εOU T , εIN , εK , εrL , εIH , εP H , εP rod , εw , επ , εGap , εrS ) .
5 Estimation This section documents and discusses the results from the estimation process of the fully speci ed,
13
linear model as set up by equations (6)-(16). Hansen's
J
statistic (χ
2
The model is well speci ed as indicated by the
(165) = 106.78, p = 0.9999).
An alternative set-up with a slightly modi ed
speci cation regarding the productivity dynamics is reported in appendix B.
5.1
Labour market dynamics
Labour market ows as reported in equations (M1) and (M2) of table 2 show the expected sign with respect to interest rates, wages and aggregate demand conditions.
In particular, housing Housing investment, It , has a strong dampening eect on unemployment, both lifting unemployment out ows and lowering unemployment in ows. Moreover, housing investment has a dampening eect on labour productivity (equation (M5)), con rming our initial hypothesis of lower productivity growth in the construction sector. At the same time, house price increases fuel (real) wage in ation and, therefore, deteriorate competitiveness (equation (M4)). Finally, equation (M3) is used as an observation equation, necessary for some of the other macro-economic relationships and in order to reduce the parameter space to be estimated.
13 Detailed
estimation results and regression output are available from the authors upon request.
13
Table 2: Labour market dynamics (M1) Unemployment in ows
INt
(M2) Unemployment out ows
OU Tt
(M3) Employment growth
∆ETt
(M4) Wage in ation
∆W agest
(M5) Productivity growth
5.2
∆P rodt
INt−1
∆LF P Rt−1
∆P rodt−1
S rt−1
T axt−1
Gapt
∆W agest−1
ItHousing
Const.
0.583*** (0.012)
-5.881*** (0.905)
-5.224*** (0.230)
0.007*** (0.001)
1.038*** (0.361)
-0.018*** (0.001)
0.340*** (0.078)
-1.215*** (0.162)
-1.909*** (0.076)
OU Tt−1
ET Rt
rtL
∆W agest
∆Kt
Gapt
Housing It−1
Const.
0.571*** (0.011)
1.194*** (0.106)
-0.007*** (0.001)
-1.641*** (0.125)
4.040*** (0.372)
0.025*** (0.001)
1.255*** (0.161)
-1.717*** (0.073)
OU Tt
INt
Const.
0.017*** (0.001)
-0.014*** (0.002)
-0.022** (0.011)
∆W agest−1
E {πt+1 }
Housing ∆Pt−1
∆P rodt−1
OU Tt
INt−1
Const.
0.647*** (0.038)
0.232*** (0.044)
3.7e-4*** (0.1e-4)
0.167** (0.072)
0.004*** (0.002)
0.007* (0.004)
-0.041** (0.020)
∆P rodt−1
∆T F Pt−1
ItHousing
Const.
0.768*** (0.054)
0.030*** (0.003)
-0.056** (0.005)
-0.023*** (0.025)
Housing and macro block
Housing investment has a strong auto-correlation component, being determined by more than 90 per cent through past investment (see equation (M6), table 3).
Financing conditions also push
up housing investment as does labour force growth. House price in ation, in turn, is positively in uenced by past housing investment as well as its own price dynamics (equation (M7)).
Table 3: Housing market (M6) Housing investment
(M7) House prices
ItHousing
∆PtHousing
Ft
Housing It−1
∆LFt−1
S rt−1
Gapt
Const.
0.004*** (0.002)
0.888*** (0.032)
0.244 (0.158)
-0.001*** (0.000)
2.3e-4* (1.2e-4)
0.009*** (0.002)
Housing ∆Pt−1
Housing It−1
Const.
0.650*** (0.023)
13.317*** (2.011)
0.878*** (0.069)
Table 4 reports the estimation results of the macro block. The Taylor rule shows important attempts to interest rate smoothing and reacts signi cantly and with the right signs to demand and in ationary shocks (see equation (M8)).
14
Real long-term interest rates are driven by overall
nancing conditions (as re ected in share prices), past long-term rates, the real (short-term) policy rate,
rtS ≡ iSt − πt ,
and government net lending, re ecting changes in the sovereign risk premiums.
Gross xed capital formation follows the dynamics as represented in equation (8) above; note that housing investment is crowding out gross xed capital formation as it enters negatively in equation (M10). The.e output gap follows the dynamics of the labour income share (see equation (M11)) and is (weakly) in uenced by government spending. Finally, the hybrid NKPC is well estimated with parameters in the expected rang (equation (M12)).
14 Note
that the estimation sample stops in 2007, and hence, the interest rate smoothing is not related to the
recent long episode of near-zero interest rates.
14
Table 4: Aggregate demand and supply (M8) Short-term rates
iS t
(M9) Real long-term rates
rtL
(M10) Gross xed capital formation
∆Kt
(M11) Output gap
GAPt
(M12) In ation rate
5.3
πt
iS t−1
GAPt−1
πt−1
Cons.
0.763*** (0.008)
0.166*** (0.008)
20.14*** (1.048)
0.682*** (0.031)
Ft−1
L rt−1
rtS
N LGQt
Cons.
-0.549*** (0.073)
0.391*** (0.009)
0.532*** (0.008)
0.201*** (0.016)
0.523*** (0.029)
Ft−1
∆GovInvt−1
∆P rodt−1
L rt−1
∆ETt−1
Housing It−1
Cons.
0.006*** (0.002)
2.022* (1.047)
0.857*** (0.103)
-0.001*** (0.000)
0.795*** (0.057)
-0.082*** (0.009)
0.023*** (0.001)
OU Tt
INt−1
∆W agest
∆Kt
GovConst−1
Housing Pt−1
Cons.
0.652*** (0.127)
-1.651*** (0.194)
8.738*** (1.360)
8.110*** (2.601)
27.952*** (2.331)
0.307*** (0.007)
-13.669*** (1.333)
πt−1
E {πt+1 }
∆T oTt−1
∆W agest−1
Cons.
0.493*** (0.014)
0.487*** (0.020)
-0.048*** (0.009)
0.022* (0.013)
-0.001** (0.001)
Fiscal block
The remainder of the estimated model covers the government accounts (see table 5).
Notice,
that government investment (equation (M14)) has a strong pro-cyclical component as it increases with accelerating employment growth. In contrast, both government spending and public revenues (equations (M13) and (M15)) show signs of automatic stabilization, the former falling with accelerating employment growth, the latter falling with an increasing in unemployment in ows and rising with unemployment out ows.
Finally, note that there is no one-to-one correspondence between
government spending and revenue ows, on the one hand, and net lending, on the other, as one-os (e.g. privatization gains) and and non-tax related revenues are not being considered here.
Table 5: Government accounts (M13) Government Consumption
(M14) Government Investment
GovConst
GovInvt
(M15) Government revenues
T axt
(M16) Net government lending
N LGQt
GovConst−1
∆ETt−1
Cons.
0.946*** (0.019)
-0.152*** (0.011)
0.012*** (0.004)
GovInvt−1
∆ETt−1
Cons.
0.920*** (0.027)
0.030*** (0.005)
0.002** (0.001)
OU Tt−1
INt
Cons.
0.008*** (0.001)
-0.005** (0.002)
0.131*** (0.011)
GovConst
GovInvt
T axt
Cons.
-76.555*** (2.298)
-155.620*** (5.469)
98.513*** (2.329)
3.602*** (0.619)
6 Simulation On the basis of the estimated model documented in the previous section, we simulate the model performing various shocks to analyse their transmission to key macroeconomic variables. In particular, we analyse the eect of a housing investment shock, a share price shock and a combined shock on unemployment and output dynamics. In section 6.2, we consider policy reforms that aect the transmission mechanism of these shocks through parametric variations. In particular, we analyse
15
the change in the reaction to these shocks when (a) wages react less to house price increases and (b) when the Taylor rule (16) is augmented to include share price dynamics.
6.1
Baseline
6.1.1 Technology, demand, labour supply and wage shocks Baseline simulations of the model dynamics with regard to technological (TFP) shocks, demand (government spending) shocks, labour supply (labour force participation) shocks and wage shocks are depicted in appendix A. The simulations reveal standard patterns of economic dynamics in our model economy, despite the fact that the estimates represent an average behaviour over 14 dierent economies.
6.1.2 A shock on the housing market The housing market can be aected both by a sudden rise in house prices or an increase in housing investment (i.e. demand for houses). Here, we rst look at a rise in housing investment (see gure 2). As the chart demonstrates, the housing boom yields a medium-term improvement on the labour market (unemployment declines for roughly 10 periods below the steady state) but then produces a long, drawn-out increase in unemployment and a drop in the output gap as both gross xed capital formation and labour productivity fall signi cantly. Importantly, the rise in real wages and the concomitant fall in labour productivity set in immediately. Hence, over the medium-term, the impact of housing investment on the labour market is dominated by the competitiveness channel rather than the wealth eect in the estimated model.
16
Figure 2: An increase in housing investment raises unemployment through the competitiveness channel
Unemployment
Output gap
Investment
0.2
0.2
0
0
0
−0.005
−0.2
10
20
30
40
−0.2
50
10
Housing investment
20
30
40
50
−0.01
0.04
0.015
0
0.02
0.01
−2
0
0.005
−4
−0.02
10 −3
6
x 10
20
30
40
0
50
Real wages
10
20
30
40
50
Real long−term interest rate 3
2
1
0
30
40
0
50
10
20
30
40
50
40
50
10
20
30
40
50
40
50
1 0.5
20
30
House prices
2
10
20
x 10Labour productivity
−6
4
0
10 −3
CPI inflation
−0.5
10
20
30
A comparison between a rise in housing investment and a rise in house prices is revealing in this regard.
Indeed, an isolated increase in house prices will have an immediate eect on unit
labour costs and dampen both employment growth and the output gap (see gure 3, panel B). In contrast, if the increase in house prices is triggered by an initial increase in housing investment, employment will rise and the output gap widen, at least initially, despite the fact that unit labour costs also deteriorate in this case (see gure 3, panel A).
Figure 3: Housing investment vs. house price increases - Labour market eects
Panel B. Rise in house prices
0.01 0.01 0.00 0.00
Deviation from steady state (in percent)
−0.00
0.00
0.40 0.20 0.00 −0.20
Deviation from steady state (in percent)
0.60
Panel A. Rise in housing investment
0
20 Employment
40 Output gap
60
80
0
Unit labour cost
20 Employment
17
40 Output gap
60
80 Unit labour cost
6.1.3 Share prices and labour markets A rise in real share prices and hence an improving in nancing conditions will reduce the unemployment rate and boost aggregate investment (see gure 4). However, the impact on the long-term performance of the economy is mitigated by the contemporaneous increase in housing investment, which also bene ts from improved nancing conditions. Over the shorter run, the improvement in aggregate investment leads to a longer boom and more lasting employment creation but the negative impact of the housing boom on competitiveness will eventually reverse the impact and lead to a decline in output and rise in joblessness.
Figure 4: Stronger asset prices improve the outlook for labour markets
5
−4 x 10 Unemployment
−4
10
0
5
−5
0
x 10
−5
Output gap 5
x 10
Investment
0
−10
10
20
30
40
50
−5
−4
2
10 −5
x 10Housing investment
6
x 10
20
30
40
50
−5
10
20
30
40
50
−5
CPI inflation 0
4
−1
2
−2
x 10Labour productivity
0
−2
10 −5
3
x 10
20
30
40
50
0
Real wages
10
20
30
40
50
−3
0.02
2
0.01
1
0
10 −3
Real long−term interest rate 5
x 10
20
30
40
50
40
50
House prices
0
0
10
20
30
40
50
−0.01
10
20
30
40
50
−5
10
20
30
6.1.4 Comparing labour market outcomes A comparison of the two types of booms shows the dierence both during the upswing period and when the recession sets in.
Panel A in gure 5 compares the unemployment impact of a
housing investment vs. a share price boom, panel B their impact on real wage growth. As the chart demonstrates, the labour market boom is longer lasting under a share price increases and the ensuing increase in joblessness much less pronounced in relative terms.
15
Partly, this can be
attributed to the fact that the impact on real wage growth is much less pronounced following a share price boom as the improvement in nancing conditions do not have a direct impact on wage claims, in contrast to housing market dynamics according to the estimation results presented in section 5.
15 Notice
the dierence in scale between the two lines.
18
Figure 5: Comparing the impact of shocks on labour market outcomes
Panel B. Real wage growth
0
10
20
Housing shock
6.2
30
40
0.000
−0.10
Deviation from steady state (in percentage points) −0.05 0.00 0.05
Deviation from steady state (in percentage points) 0.001 0.002 0.003
0.10
0.004
Panel A. Unemployment
50
0
−3
Share price shock (x10 )
10
20
Housing shock
30
40
50 −3
Share price shock (x10 )
Policies to address asset price bubbles
Sustained housing market booms and their potentially damaging long-term eects on labour markets as documented by the simulations in the previous sub-section have triggered a series of policy reactions and proposals to address them (see IMF, 2014, for an overview of measures implemented in that area following the global nancial crisis). In this sub-section, we will consider two particular (structural) policies that have been recommended to address the current economic woes: The rst one is a reduction in the pass-through of house prices on wage in ation. This is achieved by assuming that the bargaining power of workers declines (e.g. through a reduction in union density or a lowering of the strictness of employment protection legislation). In the set-up of our model, this is done through a reduction in the estimated coe cient of housing prices on wage in ation by 50 per cent. Notice that no other parameter is being changed; in particular wage rigidity is assumed to remain the same. The other policy simulation we carry out addresses the way monetary policy is being conducted. In particular, we assume that certain principles of macro-prudential regulation are being implemented in the monetary policy framework. In this simulation, we modify the Taylor rule equation (16) such that the evolution of the short-term policy interest rate is in uenced by an asset price channel. In particular, we assume an above-unity elasticity reaction of the policy rate to changes in asset prices, which are represented by an unweighted average of both share and housing prices.
6.2.1 Simulation 1: Reduction in elasticity of wages with respect to house prices Assuming a su ciently strong reduction in the elasticity of wages with respect to house prices -- in the case of gure 6, the reactivity was cut in half -- yields distinctly dierent dynamics of unemployment and the output gap, at least in the short run. Whereas housing booms created an uptick in unemployment and a large output gap in the baseline estimated model, in this policy scenario unemployment actually falls, despite the fact that the economy is still suering from a loss in competitiveness as labour productivity declines and real wages increase. The eect of the lower pass-through tapers o, however, and the reduction in labour productivity causes the output
19
gap to open up relatively quickly, with unemployment slightly moving above steady state over the medium term.
Figure 6: Lower wage pass-through of housing boom
Unemployment rate
Output gap
−3
0.2
0.2
5
0
0
0
−0.2
20 40 Housing investment
−0.2
20
40
−5
CPI inflation 0.01
0
0.02
0.005
−0.005
20
40
0
Real wages
20 40 Long−term real interest rates
−0.01
0.01
5
0.2
0.005
0
0
0
20
40
−5
20
40
20
40
Labour productivity
0.04
0
x 10 Investment
−0.2
20 40 Net government lending
20
40
6.2.2 Simulation 2: Integration of asset prices in the short-term interest rate equation When the monetary policy framework is adjusted to allow for a strong reaction of the short-term policy rate to asset prices, the dierence in short- and long-term dynamics observed in the baseline model disappears (almost) entirely (see gure 7). Unemployment remains now below steady state, even over the medium term, investment increases as well as real wages despite the fact that labour productivity continues to fall below steady state due to the loss in competitiveness resulting from strong housing investment.
The output gap remains positive over the medium term and the
government will run budgetary surpluses after the initial uptick in spending has tapered o.
20
Figure 7: Changes in the reactivity of the policy rate to asset prices
Unemployment rate 0.5
Output gap
Investment
1
0.02
0.5
0.01
0
0
0
−0.5
10 20 30 40 50
−0.5
Housing investment
10 20 30 40 50
−0.01
CPI inflation
0.04
0.04
0.02
0.02
10 20 30 40 50 Labour productivity
0 −0.005 −0.01
0
20
40
0
Real wages 0.04
20 40 Long−term real interest rates
5
−0.015
20 40 Net government lending
0.5
0 0.02
0 −5
0
20
40
−10
20
40
−0.5
20
40
7 Conclusion The paper inroduces a fully- edged housing market into an otherwise standard DSGE-cum-labour market ows model. Rather than considering only one potential pass-through, the paper introduces four dierent transmission mechanisms through which house prices and housing investment aects the macroeconomy and the labour market: A wealth eect, an aggregate demand eect, a wage eect and a competitiveness eect. Whereas the rst one is positively linked to economic activity, the other two have a depressing eect, reducing economic activity as house prices rise. A linear version of the model is brought to the data and estimated on the basis of a panel of 14 OECD countries. The estimated model is then analysed, running dierent shocks on housing investment and share prices besides the baseline supply and demand shocks. The two asset price shocks show complex dynamics whereby short- and long-term behaviour of key macroeconomic and labour market variables move in dierent directions. Both shocks generate (relatively) short-lived booms in labour market dynamics and aggregate demand that are reversed over the medium run, giving rise to long periods of joblessness. A direct comparison of the two asset price booms shows, however, that the competitiveness eect of a housing boom has particularly adverse medium-run eects, leading to smaller labour market booms and longer and more pronounced labour market busts over the longer term.
21
The paper also considers possible policy responses to these adverse eects on housing booms for the labour market. One of the suggested responses in the aftermath of the global nancial crisis was a continuation of structural policies in order to increase an economy's resilience to shocks (see, for instance, OECD, 2010). Here, we introduced such policy reforms assuming that the pass-through of house price shocks on real wage growth would be lowered as, for instance, bargaining power of trade unions decline. Our results show that such a policy shift can indeed alter the labour market dynamics following an asset price shock. However, the impact on aggregate demand will change only moderately and the output gap will eventually open up due to the still signi cant drop in labour productivity. An alternative package of policy reforms that has been suggested concerns the introduction of macro-prudential policies, intended to limit the asset price boom in the rst place (see, for instance, the recent, comprehensive overview over the issues by BIS, 2016). In the framework of our model, we introduce such a policy shift by considering a direct impact of asset price in ation on the short-term policy rate. An acceleration in asset prices will, therefore, lead to a much swifter reaction of the policy rate than under a normal in ation-targeting monetary policy framework. The simulation of such a policy shift demonstrate that this, indeed, can provide an eective answer to the impact of asset price cycles for both the labour market and aggregate demand. Unemployment does no longer overshoot over the medium-term and the output gap remains rmly positive (i.e. output runs above potential).
Over the longer term, also budget surpluses are being generated
that mitigate the initial acceleration in government spending. The approach pursued in this paper oers a new way for macro-economic modelling, setting up a dynamic general equilibrium framework suitable for being directly estimated using macro-economic panel data. The resulting simulations con rm the validity of the methodological framework and demonstrate the potential for analysing the dynamic implications of dierent policy options. Further research along the approach taken in this paper should focus on estimating the model using random-coe cient approaches in order to be able to distinguish country-speci c model dynamics and policy respones.
22
References Askenazy, P. 2013.
Capital prices and eurozone competitiveness dierentials, IZA Discussion Paper
7219 (Bonn, Intitute for the Study of Labor).
Institutional determinants of worker ows: A cross-country/cross-
Bassanini, A. et al. 2010.
industry approach, Social, Employment and Migration paper 107 (Paris, Organization for Economic Co-operation and Development (OECD)).
Bernanke, B.; Gertler, M.; Gilchrist, S. 1999. The nancial accelerator in a quantitative business cycle framework, in
Handbook of Macroeconomics
(eds. J. Taylor; M. Woodford) (Elsevier
Science). BIS (ed.) . 2016.
Macroprudential policy, BIS Papers (Basel, BIS).
Blanch ower, D.G.; Oswald, A.J. 2013.
Does high home-ownership impair the labor market?, Work-
ing Paper 13-3 (Washington, DC., Peterson Institute for International Economics). Boone, L.; Girouard, N. 2002. The stock market, the housing market and consumer behaviour, in
OECD Economic Studies, vol. 35, pp. 175200.
Branch, W.; Petrosky-Nadeau, N.; Rocheteau, G. 2016. Financial frictions, the housing market, and unemployment, in
Journal of Economic Theory, vol. 164, pp. 101135.
Coping with ageing: A dynamic approach to quantify the impact of alternative policy options on future labour supply in OECD countries, Working
Burniaux, J.M.; Duval, R.; Jaumotte, F. 2003.
Paper 371 (Paris, OECD). Calza, A.; Monacelli, T.; Stracca, L. 2013. Housing nance and monetary policy, in
the European Economic Association, vol. 11, no. S1, pp. 101122. Carlsson, M.; Eriksson, S.; Gottfries, N. 2006.
Journal of
Testing theories of job creation: Does supply create
its own demand?, Discussion Paper 2024 (Bonn, Institute for the Study of Labor (IZA)). Chakrabarti, R.; Zhang, J. 2015. Unaordable housing and local employment growth, in
Studies, vol. 52, no. 6, pp. 11341151. Corradin, S.; Fontana, A. 2013.
Urban
House Price Cycles in Europe, ECB Working Paper 1613 (Frank-
furt/M., ECB). Dai, Q.; Philippon, T. 2005.
Fiscal policy and the term structure of interest rates, Working Paper
11574 (Cambridge, MA., NBER). Davis, M.A.; Heathcote, J. 2005. Housing and the business cycle, in
Review, vol. 46, no. 3, pp. 751784.
International Economic
Davis, S.J.; Faberman, R.J.; Haltiwanger, J. 2012. Labor market ows in the cross section and over time, in
Journal of Monetary Economics, vol. 59, pp. 118.
Elsby, M.; Hobijn, B.; Sahin, A. 2013. Unemployment dynamics in the OECD, in
Economics and Statistics, vol. 95, no. 2, pp. 530548. 23
Review of
Erceg, C.; Henderson, D.; Levin, A. 2000. Optimal monetary policy with staggered wage and price contracts, in
Journal of Monetary Economics, vol. 46, pp. 281313.
Ernst, E.; Mittnik, S.; Semmler, W. 2016a. Interaction of labor and credit market in growth regimes: A theoretical and empirical analysis, in
Economic Notes.
Ernst, E.; Rani, U. 2011. Understanding unemployment ows, in
Policy, vol. 27, no. 2, pp. 268294.
Oxford Review of Economic
Ernst, E.; Semmler, W. 2010. Global dynamics in a model with search and matching in labor and capital markets, in
Journal of Economic Dynamics and Control, vol. 34, no. 9, pp. 16511679.
Ernst, E. et al. 2016b.
Predicting wages around the globe,
Working Paper (Geneva, International
Labour Organisation). Estevão, M.; Tsounta, E. 2011.
Has the Great Recession raised U.S. structural unemployment?,
Working Paper 11/105 (Washington, DC., IMF). Fang, H. et al. 2015.
Demystifying the chinese housing boom,
Working Paper 21112 (Cambridge,
MA., NBER). Farber, H.S. 2012. Unemployment in the Great Recession: Did the housing market crisis prevent the unemployed from moving to take jobs?, in
Proceedings, vol. 102, no. 3, pp. 520525. Flaschel, P.; Franke, R.; Semmler, W. 1997.
and growth in monetary economies
American Economic Review: Papers and
Dynamic macroeconomics. Instability, uctuations,
(Cambridge, MA, MIT Press).
Head, A.; Lloyd-Ellis, H. 2012. Housing liquidity, mobility, and the labour market, in
Economic Studies, vol. 79, no. 4, pp. 15591589.
Review of
Hoon, H.T. 2010. Macroeconomic eects of over-investment in housing in an aggregative model of economic activity, in
Capitalism and Society, vol. 5, no. 2.
Hughes, G.; McCormick, B. 1987. Housing markets, unemployment and labour market exibility in the UK, in
European Economic Review, vol. 31, no. 3, pp. 615641.
Iacoviello, M. 2005. House prices, borrowing constraints and monetary policy in the business cycle, in
American Economic Review, vol. 95, no. 3, pp. 739764.
. 2010. Housing in DSGE models: Findings and new directions, in
A macroeconomic perspective
Housing markets in Europe:
(eds. O.d. Bandt; T. Knetsch; J. Peñalosa; F. Zollino) (Springer),
pp. 316. . 2011.
Housing wealth and consumption, International Finance Discussion Papers 1027 (Wash-
ington, DC., Federal Reserve Board). Iacoviello, M.; Neri, S. 2007.
the US economy,
The role of housing collateral in an estimated two-sector model of
Working Papers in Economics 412 (Boston, MA., Boston College Economics
Department).
24
. 2010. Housing market spillovers:
Evidence from an estimated DSGE model, in
Economic Journal: Macroeconomics, vol. 2, pp. 125164.
American
Iacoviello, M.; Pavan, M. 2013. Housing and debt over the life cycle and over the business cycle, in
Journal of Monetary Economics, vol. 60, pp. 221238.
IMF. 2014.
World economic outlook: Legacies, clouds, uncertainties
(Washington, DC., IMF),
chap. Box 1.1.: Housing markets across the globe: An update. Jordà, O.; Schularick, M.; Taylor, A.M. 2016. The great mortgaging: housing nance, crises and business cycles, in
Economic Policy.
Judd, J.P.; Rudebusch, G.D. 1998. Taylor's Rule and the Fed: 1970-1997, in
Bank of San Francisco Economic Review, vol. 3, pp. 316. Justiniano, A.; Primiceri, G.E.; Tambalotti, A. 2015.
Federal Reserve
Credit supply and the housing boom, Working
Paper 20874 (Cambridge, MA., NBER). Knoll, K.; Schularick, M.; Steger, T. Forthcoming. No price like home: 1870-2012, in
Global house prices,
American Economic Review.
Kumhof, M.; Rancière, R.; Winant, P. 2015. Inequality, leverage, and crises, in
nomic Reivew, vol. 105, no. 3, pp. 12171245.
American Eco-
Can non-interest rate policies stabilise housing markets? Evidence from a panel of 57 economies, BIS Working Papers 433 (Basel, Bank for International Settle-
Kuttner, K.N.; Shim, I. 2013.
ments). Lee, T.H.M.; Chen, W. 2015.
Is there an asymmetric impact of housing on output?,
SFB 649
Discussion Paper 2015-020 (Berlin, Humboldt University). Liu, Z.; Miao, J.; Zha, T. 2013.
Land prices and unemployment,
NBER Working Paper 19382
(Cambridge, MA., National Bureau of Economic Research). Mankiw, N.G.; Reis, R. 2002. Sticky information versus stricky prices: A proposal to replace the New Keynesian Phillips Curve, in
The Quarterly Journal of Economics,
vol. 117, no. 4, pp.
12951328. OECD. 2010.
Going for growth. Economic policy reforms 2010
(Paris, Organization for Economic
Cooperation and Development (OECD)). Phelps, E. 1998.
assets
Structural slumps: The modern equilibrium theory of unemployment, interest, and
(Cambridge, MA, Harvard University Press).
Pissarides, C. 2000.
Equilibrium unemployment theory
(Cambridge, MA, The MIT Press).
A lost generation? Education decisions and employment outcomes during the U.S. housing boom-bust cycle of the 2000s, Working Paper 1892 (Frankfurt/M., ECB).
Popov, A.; Laeven, L. 2016.
Rajan, R.G. 2010.
Fault lines: How hidden fractures still threaten the world economy
NJ, Princeton University Press).
25
(Princeton,
Rünstler, G.; Vlekke, M. 2016.
Business, Housing and Credit Cycles, Working Paper 1915 (Frank-
furt am Main, ECB). Rupert, P.; Wasmer, E. 2012. Housing and the labor market: Time to move and aggregate unemployment, in
Journal of Monetary Economics, vol. 59, no. 1, pp. 2436.
Shimer, R. 2012. Reassessing the ins and outs of unemployment, in
Review of Economic Dynamics,
vol. 15, pp. 127148. Woodford, M. 2001. The Taylor Rule and optimal monetary policy, in
Reivew, vol. 91, pp. 232237. Wu, T. 2003. What makes the yield curve move?, in
Economic Letter, vol. 2003-15. Zoega, G. 2009.
Employment and asset prices,
American Economic
Federal Reserve Bank of San Francisco
Working Paper 46 (Reykjavik, Central Bank of
Iceland).
26
Appendix A: Summary statistics Table 6: Summary statistics Country
stats
∆P rodt
N LGQt
∆PtHousing
ItHousing
Australia
mean
6.189
72.62
6.192
1.48
-1.636
3.195
7.799
3.072
8.981
sd
2.609
2.611
4.138
1.387
2.341
5.902
0.7465
3.621
4.016
min
1.403
69.4
0.2504
-1.738
-5.478
-6.104
6.45
-6.259
4.747
max
10.82
77.73
15.3
3.673
2.045
17.19
9.598
9.613
17.61
mean
7.978
73.65
4.757
1.093
-2.861
2.717
8.494
3.534
7.502
sd
1.958
4.966
3.367
1.095
3.611
6.184
1.029
2.461
3.519
min
4.366
63.9
0.1656
-0.7833
-9.126
-13.96
6.552
-2.127
2.356 18.07
Canada
France
Germany
Japan
New Zealand
80
12.47
3.189
2.945
14.99
10.2
8.405
68.59
5.14
2.028
-2.163
2.734
8.649
3.429
7.553
sd
3.178
0.8501
4.218
1.194
1.767
5.135
1.759
2.315
3.591
min
2.294
66.95
0.5417
-0.0176
-6.424
-5.984
6.653
-2.647
2.106
max
12.07
70.29
13.65
5.344
0.7043
13.33
12.14
6.701
15.26
mean
5.939
75.6
3.051
1.334
-2.818
-0.3141
7.582
3.294
5.729
0.9179
2.221
2.377
1.002
1.026
2.676
-0.1294
0.3273
-9.667
-4.569
6.08
1.796
2.106
11.15
79.59
7.032
3.356
1.314
5.201
8.936
5.135
12.14
66.93
5.902
3.368
-4.084
4.535
7.099
3.25
7.075
sd
4.517
2.737
5.267
1.824
4.939
7.154
0.9657
3.281
3.902
min
3.67
63.36
1.417
-0.1569
-13.29
-6.722
5.352
-3.106
2.106
max
16.98
74.16
20.37
6.543
4.8
24
8.958
8.095
14.32
mean
8.762
60.05
7.845
1.836
-7.003
2.515
8.772
2.375
9.953
sd
2.282
1.314
6.019
1.945
3.6
10.07
2.226
4.2
5.301
min
5.279
57.55
1.663
-2.124
-12.38
-10.97
6.588
-9.548
2.106
max
11.83
63.03
21.06
5.586
-0.8632
38.83
14.77
7.995
19.91
mean
2.796
74.35
3.192
2.666
-2.914
0.4988
5.44
1.857
4.599
sd
1.235
3.093
4.79
2.515
3.35
6.29
1.471
3.603
3.656
min
1.118
70.19
-0.8953
-1.404
-11.16
-13.96
2.953
-15
0.029
max
5.367
80.33
23.18
11.09
2.05
19.61
8.367
5.921
14.51
mean
4.167
65.66
7.264
1.337
0.6633
3.814
9.207
1.901
10.82
mean
max
United States
1.959
73.03
9.426
min
United Kingdom
2.304
max
sd
Sweden
3.068 0.5556
mean
max
Spain
iSt
11.95
min
Portugal
rtL
7.937
sd
Norway
πt
mean
min
Italy
LF P Rt
max
sd
Ireland
Ut
2.991
1.259
5.719
2.328
3.607
9.008
1.588
4.667
4.82
0.0831
63.3
-0.1143
-6.768
-6.336
-9.713
7.418
-8.565
4.828
10.3
68.45
17.15
6.591
5.929
31.81
13.73
7.508
23.31
3.088
77.12
5.295
2.274
5.549
2.896
7.111
2.959
8.793
1.521
3.487
3.576
1.547
4.895
7.372
2.071
3.111
4.185
0.7614
68.57
0.4655
-0.9178
-1.852
-11.35
3.678
-4.401
2.007 15.37
5.96
81.1
13.66
4.599
18.48
22.53
10.17
7.26
mean
5.725
70.63
11.82
2.109
-5.147
0.3759
9.627
1.442
9.998
sd
2.062
4.676
8.865
3.533
1.962
3.497
0.9099
6.893
6.665
min
1.687
59.74
2.294
-8.608
-8.883
-5.254
7.383
-19.15
2.106
max
8.529
78.25
33.06
12.05
-2.65
6.756
11.56
11.18
24.9
mean
12.46
61.64
8.343
2.073
-2.429
4.716
11.22
1.691
10.19
sd
7.076
5.082
5.899
2.102
2.713
9.664
2.193
4.77
5.686
min
1.419
56.18
1.834
-0.4593
-7.321
-9.795
8.457
-14.21
2.106
max
23.88
73.09
24.54
7.9
2.215
33.51
14.98
7.956
20.05
mean
4.445
70.56
5.481
1.934
-0.067
1.748
9.695
3.312
7.544
sd
2.797
2.17
4.019
1.457
4.546
6.943
4.23
2.77
4.176
min
1.508
65
-0.2671
-1.414
-11.17
-15.66
3.416
-1.873
1.717
max
10.19
73.96
13.71
4.81
6.502
11.32
15.98
7.789
14.17
mean
6.419
75
6.398
2.003
-2.734
4.782
10.13
2.749
8.924
sd
2.962
1.191
5.764
1.45
2.592
9.908
1.925
3.698
3.425
min
2.041
72.97
0.7853
-1.656
-7.978
-14.38
7.686
-11.03
3.666
max
11.81
77.43
24.21
5.54
3.675
26.26
14.34
6.707
16.62
mean
5.916
64.51
4.672
1.272
-2.777
2.07
7.096
2.839
7.081
sd
1.429
2.447
2.968
1.24
1.951
3.012
1.111
2.581
3.372
min
3.412
59.62
1.552
-2.433
-5.769
-3.523
4.969
-3.497
1.171
max
9.545
67.1
13.55
3.617
1.621
7.392
10.04
8.138
16.83
27
Appendix B: Alternative speci cation Table 7 presents and alternative estimation speci cation of equations (6)-(16), where labour productivity is replaced by TFP growth in equation (A1). Looking through the estimated coe cients, this speci cation seems less well identi ed as has, therefore, not been used for the simulations in section 6. All variable names as de ned in section 3.
Table 7: Alternative speci cation - TFP growth - Part I
INt−1 LF P Rt−1 ∆wt ∆T F Pt−1 S rt−1
T axt−1 Gapt ItHousing
(A1) In ow 0.628*** (0.0119) 0.829 (0.820) 0.391*** (0.0529) 0.415*** (0.0172) 0.00559*** (0.000513) 1.916*** (0.178) -0.0190*** (0.00128) -1.293*** (0.117)
(A2) Out ow
(A3) MPolicy
ET Rt rtL ∆Kt Housing It−1
(A5) Housing
(A6) GovCons
0.804*** (0.00915)
-0.000582*** (4.27e-05)
0.0263*** (0.00108)
0.000239** (0.000117)
0.00766*** (0.00119)
-0.0845*** (0.00818)
0.889*** (0.0280)
0.00519*** (0.00149) 2.060** (0.872) 0.928*** (0.0964) -0.00128*** (0.000182) 0.803*** (0.0508)
0.00459*** (0.00153)
-0.0619*** (0.00746) 11.04*** (0.490)
πt−1 Ft−1 GovInvt−1 ∆P rodt−1 L rt−1
∆ETt−1
0.946*** (0.0144) -0.147*** (0.0110)
GovConst−1 ∆ETt OU Tt INt
Observations
(A8) Taxes
-1.610*** (0.115)
Gapt−1
Constant
(A7) Employment
0.227 (0.159)
0.568*** (0.00996) 1.181*** (0.0961) -0.00733*** (0.00108) 3.828*** (0.190) 1.247*** (0.165)
OU Tt−1
(A4) Investment
-2.314*** (0.0653) 247
-1.705*** (0.0648) 247
0.0971*** (0.0357) 247
0.0224*** (0.00123) 247
0.00954*** (0.00194) 247
0.0124*** (0.00284) 247
0.0165*** (0.00113) -0.0158*** (0.00191) -0.0298*** (0.0104) 247
Note: Standard errors in parentheses; *** p
