House prices, monetary policy and regional heterogeneity

Transcription

1 House prices, monetary policy and regional eterogeneity Knut Are Aastveit André K. Anundsen First Version: April, 2016 Abstract Te effectiveness of monetary policy in affecting ouse prices depend bot on te nature of te sock; expansionary versus contractionary, and on local ousing market caracteristics. In particular, our results suggest tat monetary policy is more effective wen it is expansionary and in markets wit low ousing supply elasticities. Wile expansionary and contractionary socks ave similar impacts on ouse prices in markets wit an elastic ousing supply, te effect of expansionary socks are markedly larger tan te impact of contractionary socks in supply inelastic areas. Tese conclusions are drawn based on an empirical examination of te effects of exogenous monetary policy socks on ouse prices using local projection metods on a panel of te 100 largest metro areas in te US over te period 1980Q1 2008Q4. Keywords: House prices; Heterogeneity; Monetary policy; Non-linearity; Supply elasticities; JEL classification: E32, E43, E52, R21, R31 Tis paper sould not be reported as representing te views of Norges Bank. Te views expressed are tose of te autors and do not necessarily reflect tose of Norges Bank. We tank Jean Imbs, Kevin Lansing, Juan Rubio-Ramirez and seminar participants at Norges Bank and Banque de France for useful comments. Norges Bank, Knut-Are.Aastveit@norges-bank.no Norges Bank Andre-Kallak.Anundsen@norges-bank.no 1

2 1 Introduction We study te asymmetric effect of exogenous monetary policy socks on ouse prices using a panel of 100 US MSAs over te period 1980Q1-2007Q4. In our baseline exercise, we measure te monetary policy sock, using an updated version of te Romer and Romer (2004) narrative sock series, provided by Wieland and Yang (2016) and estimate impulse responses using panel data and local projection metods as in Jordà (2005). We ask te following two questions: (i) Is te effect of monetary policy dependent on local ousing market conditions? (ii) Is te impact of monetary policy asymmetric? We start by documenting a strong role of monetary policy in influencing aggregate ouse prices in te US. Following a one percentage point contractionary monetary policy sock aggregate ouse prices decrease wit about 8 percent after tree to four years. Tese estimates are in line wit te rapidly growing literature investigating te nexus between monetary policy and ouse prices (see e.g., Del Negro and Otrok (2007), Glaeser et al. (2012), Goodart and Hofmann (2008), Iacoviello (2005), Jarocinski and Smets (2008), Jordà et al. (2015) and Williams (2011, 2015)). Tat said, an aggregate investigation masks te major eterogeneities existing across regional US ousing markets and national ouse price cycles are often driven by developments in certain regional markets, see e.g., Capozza et al. (2004), Glaeser et al. (2008) and Malpezzi and Wacter (2005) and Saiz (2010). For instance, wile ouse prices increased by more tan 160% in some coastal areas of Florida and California from 2000 to 2006, tey increased by less tan 20% in inland open space areas of te Midwest. Te presence of eterogeneous regional socks witin a country may pose a callenge to monetary policy making. A key question is weter monetary policy sould pay attention to tese regional disparities? Te fact tat monetary policy is common across regional markets may suggest tat canges in monetary policy could ave similar effects across member regions. However, on te oter side, differences in regional economic conditions could impact te transmission of monetary policy, potentially amplifying regional 2

3 eterogeneity. Many ave for instance wondered weter te low interest rate environment tat prevailed in te years before te 2008 crisis contributed to te ouse price booms experienced in many Western economies before te recent cras. Oters, suc as Mian and Sufi (2009) and Favara and Imbs (2015) ave empasized te role of lax lending standards associated wit securitization. A branc of te literature ave attributed regional variations in ouse price dynamics to eterogeneous supply side restrictions see e.g., Malpezzi (1996), Green et al. (2005), Saiz (2010), Gyourko et al. (2008), Glaeser (2009), and Anundsen and Heebøll (2016), Huang and Tang (2012) and Glaeser et al. (2008) sow tat ouse price booms tend to be larger in markets wit an inelastic ousing supply. By adding MSA-dependent ousing supply elasticities, as calculated by Saiz (2010), as an interacting variable in our model, we still find a strong role of monetary policy in affecting ouse prices. However, consistent wit te cross section studies alluded to above, te eterogeneoty in te transmission of monetary policy is considerable. For instance, wereas te cumulative drop in ouse prices following a one percentage point contractionary monetary policy sock is estimated to be more 11 percent in Miami (FL) after four years, ouse prices in Dayton (OH) are predicted to fall less tan tree percent over te same period. Motivated by recent findings of an asymmetric effect on output from contractionary and expansionary fiscal policy socks (see Auerbac and Gorodnicenko (2012) and Owyang et al. (2013)) and monetary policy socks (see Angrist et al. (2013) and Tenreyro and Twaites (2016)), our final exercise is to explore asymmetric effects on ouse prices of contractionary and expansionary socks. We find evidence tat expansionary socks ave a greater impact on ouse prices tan contractionary socks, wic could pose a callenge from a financial stability point of view. Tere are nuances to tis finding, owever. We find tat te increase in ouse prices following an expansionary sock is twice as ig (in absolute value) as te fall in prices following a contractionary sock in Miami (FL) and San Francisco (CA) bot markets wit an inelastic ousing supply. For Dayton (OH), an area wit an elastic ousing supply, te effect of contractionary socks 3

4 are stronger. Tus, unless te ousing supply elasticity is very ig, monetary policy as a far larger impact on fueling tan taming ouse price growt, wic may be particularly important to keep in mind wen ten tere are trade-offs between developments in economic activity, inflation and ouse prices. Our investigation of te asymmetric transmission of monetary policy socks to ouse prices continue as follows. Te next section sketces a simple skeleton model tat gives a teoretical foundation for our empirical investigation. Section 3 presents te data we utilize. Section 4 documents our empirical findings on te eterogenous and asymmetric effects of monetary policy. Te final section concludes. 2 Teoretical motivation Following Glaeser et al. (2008), we consider an economy consisting of several eterogeneous ousing markets wit different supply elasticities. Specifically, some regions are open space areas wit no regulations on building permits, wile oter regions are naturally restricted, e.g. by mountains or water, or by te local regulatory framework. In particular, we will sed ligt on possible eterogeneities and asymmetries in te response to expansionary and contractionary monetary policy socks. In eac period, te law of motion of capital accumulation for area i is given as: 1 H i,t = H i,t 1 + I i,t (1) were H i,t is te ousing stock at time t, wile I i,t represents new investments in ousing capital. We assume tat investments are determined according to Tobin s Q teory (Tobin, 1969), i.e. new construction projects are initiated as long as te market price, P H i,t, exceeds te marginal cost of construction, MC i,t. Wen considering eterogeneous areas of different sizes, te number of new construction projects initiated in eac period will naturally depend on te size of te market in 1 For now, we abstract from depreciation of te existing stock. 4

5 question. To take account of tis, we assume tat te marginal cost of investments is inversely proportional to te existing ousing stock, i.e. tere is a larger construction capacity in bigger markets. Te marginal cost function for area i takes te following form: MC i,t (I i,t ) = C 0,i (I i,t /H i,t 1 + 1) 1/ϕ i, ϕ i > 0 i were ϕ i is te time invariant area specific supply elasticity, wile C 0,i is a positive variable measuring fixed costs of ousing construction (we disregard time varying construction costs for now). Setting te price equal to te marginal cost, we get te following investment function: { I i,t = H i,t 1 max 0, ( ) ϕi P Hi,t 1} C 0,i (2) Given a non-zero supply elasticity, it follows from 2 tat tere will be positive investments if and only if prices exceed te fixed costs of construction. Moreover, te size of te investment response depends on bot te size of te market (as measured by H i,t 1 ) and te supply elasticity. Te two extreme cases are interesting: In a completely supply elastic market (ϕ i ), a positive price-to-cost ratio implies tat investments become infinite, wile in a market were supply is completely inelastic (ϕ i 0), investments will be zero and independent of ouse prices. From (1) and (2), we find tat a log transformation (lower case letters) of te supply equation yields: 2 i,t = i,t 1 + max {0, ϕ i (p i,t c 0,i )} (3) It follows tat te log supply curve will be piecewise linear and kinked; only if te price exceeds te fixed cost of construction will supply increase as a function of te supply elasticity, ϕ i, and te price-to-cost ratio (Tobin s Q). Hence, supply is assumed completely { ( ) 2 ϕi } P Hi,t Tis is seen by rewriting (1) using (2); H i,t = H i,t 1 max 1, C 0,i and ten taking logs. 5

6 rigid downwards, motivated by te fact tat ousing is usually neiter demolised nor dismantled (see also te discussion in Glaeser and Gyourko (2005)). We follow custom wen it comes to te modelling of te demand side. For eac area, it is assumed tat demand is determined in accordance wit te life cycle model of ousing, see e.g. Meen (1990, 2001) and Muellbauer and Murpy (1997), and te references terein. For area i, a logaritmic representation of te inverted demand function is given as: p i,t = v 0,t i t + v 1,i,t + v 2 i,t, v 0, v 2 < 0 (4) were te v 0 is te semi-elasticity of ouse prices wit respect to te interest rate, v 1,i,t measures oter demand sifters, suc as income and migration. Te parameter v 2 measures te price elasticity of an increase in te number of ouses (te inverse demand elasticity). Let us assume tat market i initially is in equilibrium (p i,t = c 0,i ), wic also implies tat I i,t = 0 and ence tat H i,t = H i,t 1. Ten, market i is it by an expansionary monetary policy sock. Tis will lead to a greater increase in ouse prices in more inelastic markets. Figure 1 illustrates tis results for te case of two markets wit different supply elasticities (elastic and inelastic). In tis figure, te sort-run supply curve at time t is drawn as a vertical line due to te fact tat increasing te ousing stock is not done over nigt. Te original long-run supply curve as a kink at point A te market equilibrium. Tis is to capture te implication implied by our stylized model, namely tat ouses are durable and tat once tey are built, tey are usually not destroyed. As seen, a positive demand sock (D t to D t+1 ) primarily leads to quantity adjustments in supply elastic markets, wile te sock is mostly absorbed in terms of iger prices in inelastic markets. To ensure market clearing, te part of te adjustments tat as to be made in te form of iger prices will be larger te lower te supply elasticity. Tus, as expected, te conjecture of a standard supply-demand story is tat expansionary sock 6

7 as a greater impact on ouse prices te lower is te elasticity of supply. At te same time, te new sort run supply curve will sift, leading to a new kink in te long-run supply curve at point B. Tus, te dotted part of te old long-run supply curve is no longer relevant, since we assume tat te new ouses tat are built will not be destroyed. Tus, a negative sock would lead to an adjustment along te vertical part of te supply curve. Figure 1: Expansionary monetary policy in supply-elastic vs. - inelastic markets p SSR t SSR t p S t SSR t+1 SLR SLR p t+1 B p t+1 p t = c 0 A B p t = c 0 A D t+1 D t+1 t t+1 D t t t+1 D t Market 1: Elastic supply Market 2: Inelastic supply Note: D t is te original demand curve, wile D t+1 is te demand curve after te expansionary monetary policy sock. SSR t is te original sort-run supply curve and SSR t+1 is te sort-run supply curve after te sock materializes. Te long-run supply curve is given by SLR. In Figure 2, we consider a contractionary monetary policy sock. Since supply is rigid downwards, tis means tat te demand curve sifts along a vertical supply curve, independent of te supply elasticity. Te conjecture is terefore tat te drop in ouse prices following a contractionary monetary policy sock is independent of te supply elasticity. Furtermore, te drop in prices following te contractionary sock will always be greater (in absolute value) tan te increase in ouse prices following a similar sized expansionary sock at least as long as supply is not completely inelastic, in wic te supply curve would always be vertical. 7

8 Figure 2: Contractionary monetary policy in supply-elastic vs. - inelastic markets p SSR t p S t SLR SLR p t = c 0 A p t = c 0 A p t+1 B p t+1 B t = t+1 D t+1 Market 1: Elastic supply D t t = t+1 D t+1 Market 2: Inelastic supply Note: D t is te original demand curve, wile D t+1 is te demand curve after te expansionary monetary policy sock. SSR t is te original sort-run supply curve and SSR t+1 is te sort-run supply curve after te sock materializes. Te long-run supply curve is given by SLR. D t Notes to self: Section is incomplete, but we sould: 1. Extend te model to allow for endogenous acceleration in demand in upturns (expectations and credit availability) Can explain wy expansionary socks ave a greater impact tan contractionary sock in inelastic markets 2. Consider modeling flow of supply instead of te ousing stock, i.e. wat is available for sale at any point in time is te sum of new investments and a fraction of te existing stock. Mecanisms will be similar, but tis is more in line te approac taken in e.g., Glaeser et al. (2008) 3. Derive results analytically in addition to te figures 4. Consider dropping te figures 8

9 3 Data and descriptive statistics 3.1 Data Our data set includes te 100 largest Metropolitan Statistical Areas (MSAs) in te United States, covering about 60 percent of te entire US population and all but four of te 50 US states. 3 Following te Census Bureau, te US may be split into four distinct regions: West, Sout, Midwest and Norteast, confer Figure 3. Wit reference to tose regions, Figure 3: Main geograpical regions in te US our data set includes 25 areas in te West and te Midwest regions, wile we ave 20 MSAs situated in te Norteast and 30 in te Sout. In addition to aving a ric crosssectional dimension, we also ave a fairly long time series dimension for eac of tese areas. Te sample runs troug te period from 1980q1 to 2010q2 (T = 122) for 82 of te areas, wile te sortest samples (Fargo (ND-MN) and Sioux Falls (SD)) contain 95 observations. Tus, te sample covers bot te recent ousing cycle and te previous boom-bust cycle in te period for a majority of te areas considered. 4 3 Note tat some of te MSAs belong to multiple states. 4 Here, we rely on te boom-bust cycle classification provided by Glaeser et al. (2008). 9

10 Te ouse price data ave been gatered from te Federal Housing Finance Agency (FHFA), wile ouseolds disposable income, te ousing stock and te CPI index used for te nominal-to-real transformations ave been supplied by Moody s Analytics. We combine te MSA specific data wit te Romer and Romer (2004) narrative monetary policy sock series. Romer and Romer propose a novel procedure to identify monetary policy socks. First, tey use te narrative approac to extract measures of te cange in te Fed s target interest rate at eac meeting of te Federal Open Market Committee (FOMC) between 1969 and Tey ten regress tis measure of policy canges on te Fed s real-time forecasts of past, current, and future inflation, output growt, and unemployment. Te residuals from tis regression constitute teir measure of monetary policy socks. Te Romer and Romer sock series as gained acceptance as an exogenous indicator of monetary policy socks and as been widely used to study te transmission of monetary policy socks, see e.g. Coibion (2012), Ramey (2016). We use an updated version of te Romer and Romer (2004) narrative sock series, provided by Wieland and Yang (2016). To account for regional eterogeneities, we use te MSA specific supply elasticities calculated by Saiz (2010) based on topograpic measures of undevelopable land. 4 Monetary policy socks and ouse prices 4.1 Monetary policy socks and ouse prices Te starting point for our empirical analysis is to encompass te findings of a strong effect of monetary policy on ouse prices, as documented in te literature. Our modus operandi is te local projection framework suggested in Jordà (2005). We use tis framework to estimate te cumulative percentage response to ouse prices quarters after a monetary policy sock, letting run from 0 to 20. Te advantage of using te local projection approac is tat it allows us to study te various non-linearities we are interested in, wic would be vastly complicated and maybe even infeasible in a standard VAR 10

11 framework. In addition, our parameters of interest (te response in ouse prices to a monetary policy sock) are confined to one equation in te underlying VAR system, i.e., te ouse price equation. We start by considering a dynamic fixed effects model wit no eterogeneity or asymmetries: p i,t+ p i,t 1 = α,i + β RR t + Γ Z i,t 1 + ε i,t+ (5) were p i,t+ p i,t 1 is te cumulative cange in log ouse prices after -orizons, RR is te Romer and Romer (2004) sock, wile Z is a vector of control variables, including lagged canges in log ouse prices, lagged values of te log cange in disposable income per capita, lagged canges in net migration rates and lagged canges in te log of te ousing stock. For all variables, we include four lags. β measures te cumulative cange in ouse prices quarters after te monetary policy sock. Cumulative responses at = 2, 4, 8, 12 and 16 are reported in Table 1 for a monetary policy sock tat raises te interest rate by one percentage point. In te same table, we report cumulative effects on ouse prices for te same orizons following a sock to disposable income per capita of one percent. It is evident tat an exogenous increase in te Fed funds rate as a negative effect on ouse prices. Wile ouse prices are predicted to fall by a little more tan one percent after 2 quarters, te fall is substantial just below 8 percent after four years. Wile sizeable, te effects we find are very muc in line wit estimates documented in te literature on international and US data (see Williams (2015) for an excellent overview). We also see tat our results imply tat ouse prices increase by close to one percent wen income increases by unity implying a constant price-to-income ratio four years after income is permanently increased by one percent. In Figure 4, we plot te cumulative effect on ouse prices for all orizons running from 0 to 20. Te figure sow te same picture as we can read from te table, namely tat an 11

12 Table 1: Effect of monetary policy sock on ouse prices, symmetric and omogenous response. =2 =4 =8 =12 =16 MP sock (0.16) (0.20) (0.36) (0.56) (0.74) Income per cap (0.04) (0.05) (0.08) (0.11) (0.15) MSA fixed effects Yes Yes Yes Yes Yes Observations MSAs Witin R Notes: Te dependent variable are te cumulative log canges in te FHFA ouse price index at orizon = 2, 4, 8, 12 and 16. Results are based on estimating equation 5 using a fixed effect estimator and te data set cover a panel of 100 US MSA s countries over te period 1980q1 2007q4. Standard errors are clustered by MSA and reported in absolute value in parentesis below te point estimates. Te asterisks denote significance levels: * = 10%, ** = 5% and *** = 1%. exogenous contractionary monetary policy sock as a sizeable and negative impact on ouse prices. 12

13 quarters Figure 4: Response on te aggregate ouse price from a 1 percentage point contractionary monetary policy sock. 4.2 Asymmetric effects of monetary policy Te durability of ousing implies tat te responses to contractionary and expansionary socks may differ. In particular, if te demand responses to a contractionary and an expansionary monetary policy sock are equal (in absolute value), te simple demandsupply framework described in Section 2 suggests tat te cange in ouse prices will be larger for te contractionary sock. To investigate te relevance of tis conjecture, we consider a modified version of (5): p i,t+ p i,t 1 = α,i + β Exp. RR + t + β Cont. RR t + Γ Z i,t 1 + ε i,t+ (6) were RR + t is a variable measuring expansionary socks and is calculated as RR + t = RR t I(RR t 0), were I(RR t 0) is an indicator variable taking te value one for expansionary sock and a value of zero oterwise. Contractionnary socks are measured 13

14 by RR t = RR t (1 I(RR t 0)). Tus, β Exp. is te cumulative effect on ouse prices after quarters following an expansionary monetary policy sock, wile β Cont. measures te effect of a contractionary monetary policy sock. Estimating 6 for = 2, 4, 8, 12 and 16, we obtain te results reported in te Table 2. 5 Te contractionary sock as an expected negative impact on ouse prices, wile te expansionary sock exercises a positive impact on ouse prices. More intriguing, and in contrast to te simple demand-supply framework, we find tat expansionary socks ave a greater impact on ouse prices tan contractionary socks. In particular, wile an expansionary sock immediately leads an increase in ouse prices, it takes about a year before a contractionary sock materializes into a drop in ouse prices. We will later investigate possible explanations of tis, but one obvious explanation is tat te demand response to an interest rate increase differ from tat of a reduction in te interest rate. Figure 5 summarizes te te cumulative responses for all orizons from zero to 20 following an expansionary and a contractionary sock, and unsurprisingly te figure mirrors te results reported in Table 2. Table 2: Effects of contractionary and expansionary monetary policy socks on ouse prices. =2 =4 =8 =12 =16 Contr. MP sock (0.25) (0.29) (0.45) (0.62) (0.75) Exp. MP sock (0.23) (0.35) (0.54) (0.79) (1.06) Income per cap (0.04) (0.05) (0.09) (0.12) (0.16) MSA fixed effects Yes Yes Yes Yes Yes MSAs Witin R Observations Notes: Te dependent variable are te cumulative log canges in te FHFA ouse price index at orizon = 2, 4, 8, 12 and 16. Results are based on estimating equation 6 using a fixed effect estimator and te data set cover a panel of 100 US MSA s countries over te period 1980q1 2007q4. Te specification allows te response in ouse prices to differ depending on weter te monetary policy sock is expansionary or contractionary. Standard errors are clustered by MSA and reported in absolute value in parentesis below te point estimates. Te asterisks denote significance levels: * = 10%, ** = 5% and *** = 1%. 5 Note tat te effect of te expansionary sock is normalized to be β Exp., suc tat it picks up te effect of a reduction in te interest rate of 1 percentage point. 14

15 quarters Expansionary quarters Contractionary Figure 5: Asymmetric effect on te aggregate ouse price from a 1 percentage point monetary policy sock. Te first subfigure, labeled Expansionary, sows te response on aggregate ouse prices of a monetary policy sock tat decrease te interest rate. Te second subfigure, labeled Negative, refers to a monetary policy sock tat decrease te interest rate. 15

16 4.3 Regional variations in effectiveness of monetary policy Wile our results are consistent wit te literature documenting a strong role of monetary policy in affecting ouse prices, it is well known tat te evolution of ouse prices differ substantially across US MSAs (see e.g., (Glaeser et al., 2008; Capozza et al., 2004; Malpezzi and Wacter, 2005)). Figure 6 displays te evolution of real ouse prices for four of te MSAs contained in our data set. Te areas were cosen to illustrate four different types of ousing markets, located in different regions of te US. Real ouse prices ave moved quite differently in te four different areas, wit a muc more pronounced run-up (and subsequent bust) in San Francisco and Boston tan in Houston and Wicita over te previous decade. Figure 6: Log of real ouse prices, 1980q1-2010q2. Our teory discussion suggested tat local variations in ousing supply elasticities may lead to different ouse price responses following a demand sock. To investigate weter te supply elasticity as an impact on ow monetary policy socks are absorbed, we 16

17 consider an equation of te following form: p i,t+ p i,t 1 = α,i + β RR t + β El. Elasticity i RR t + Γ W i,t 1 + ε i,t+ (7) were Elasticity i is te time-invariant supply elasticities calculated in Saiz (2010), wit a iger value indicating a more elastic ousing supply. For a particular area, i, te cumulated response to ouse prices in period following a monetary policy sock is given as β + β El. Elasticity i. Te vector of controls include te same set of controls as previously, but also interactions of te oter demand sifters (migration and income) and te supply elasticity. Tis is because teory also suggests tat te reaction to oter demand sifters sould be lower te lower is te elasticity of supply. Table 3 summarizes results from estimating (7). We find tat an exogenous increase in te policy rate as a smaller negative effect on ouse prices at all orizons te iger is te elasticity of supply, wic is in line wit te teory model. Consistent wit tis, we also find a smaller effect on ouse prices following an increase in per capita disposable income in more supply elastic areas. We ave calculated te response for 5 different MSAs wit very different supply elasticites; Dayton, Kansas City, Scranton, San Francisco and Miami. Wile ouse prices are predicted to fall by less tan 3 percent in Dayton 4 years after a contractionary monetary policy sock, ouse prices in Miami are precited to fall by more tan 11 percent over te same orizon. Our results terefore suggest a substantial difference in ouse price responses for areas wit different supply elasticities. To sed some more ligt on tis, Figure 7 display te response in ouse prices for areas wit low, normal and ig supply elasticities. An area is classified as aving a low elasticity of supply if it is at least one standard deviation below te mean. Similarly, te elasticity is ig if it is at least one standard deviation above te mean. All oter areas 17

18 Table 3: Effect of contractionary monetary policy sock wit different supply elasticities. =2 =4 =8 =12 =16 MP sock (0.23) (0.39) (0.64) (0.96) (1.22) MP sock Elasticity (0.09) (0.16) (0.27) (0.40) (0.49) Income per cap (0.07) (0.09) (0.15) (0.20) (0.27) Income per cap. Elasticity (0.03) (0.04) (0.06) (0.07) (0.09) MSA fixed effects Yes Yes Yes Yes Yes Effects of contractionary monetary policy sock for five different MSAs: Dayton (El.=3.64) (0.21) (0.31) (0.51) (0.77) (0.96) Kansas (El.=3.19) (0.18) (0.26) (0.42) (0.64) (0.80) Scranton (El.=1.62) (0.15) (0.20) (0.34) (0.51) (0.68) San Francisco (El.=0.66) (0.19) (0.30) (0.50) (0.74) (0.96) Miami (El.=0.60) (0.19) (0.31) (0.51) (0.76) (0.98) Observations MSAs Witin R Notes: Te dependent variable are te cumulative log canges in te FHFA ouse price index at orizon = 2, 4, 8, 12 and 16. Results are based on estimating equation 7 using a fixed effect estimator and te data set cover a panel of 100 US MSA s countries over te period 1980q1 2007q4. Te specification allows te response in ouse prices to differ depending on te elasticity of supply, as calculated in Saiz (2010). Standard errors are clustered by MSA and reported in absolute value in parentesis below te point estimates. Te asterisks denote significance levels: * = 10%, ** = 5% and *** = 1%. 18

19 are classified as aving a normal elasticity. 6 Responses are remarkably different. Wile te most inelastic areas experience a large drop in ouse prices following a contractionary monetary policy sock, te most elastic areas see only a modest fall in ouse prices. 6 To compute te cumulative responses for te low, normal and ig elasticity areas, we estimated an auxiliary regression of te following form: p i,t+ p i,t 1 = α,i + β Low Eli Low RR t + β Normal Eli Normal RR t + β Hig El Hig i RR t + Γ W i,t 1 + ε i,t+ were Eli Low = El i I(El i < Ēl σ El), wit Ēl signifying te mean elasticity and σ El being te standard deviation of te elasticity measure. Likewise, El Hig i = El i I(El i > Ēl+σ El) and Eli Normal = El i I(Ēl σ El El i Ēl + σ El). 19

20 quarters Low quarters Normal quarters Hig Figure 7: Effect on ouse price from a 1 percentage point monetary policy sock for MSA s wit different ousing supply elasticity, as measured by Saiz (2010). Low elasticity refers to te lower decile of elasticity, i.e., te most inelastic markets, wile Hig elasticity refers to te decile wit te most elastic supply. Normal refers to rest. 20

21 4.4 Asymmetric effects of monetary policy wit different supply elasticities Te final prediction of te teory model is tat wile an expansionary monetary policy sock ave a greater impact on ouse prices in areas wit an inelastic ousing supply, te response to a contractionary sock sould be similar across areas, due to te downward rigidity of ousing supply. We investigate tis by allowing expansionary and contractionary socks ave different effects for different supply elasticities. We consider a specification of te following form: p i,t+ p i,t 1 = α i + β Exp. RR t + + β Exp.,El. Elasticity i RR t + + β Cont. RR t + β Cont.,El. Elasticity i RR t + ΓW i,t 1 + ε i,t (8) Wit tis specification, te cumulative response to ouse prices in area i after quarters following an expansionary monetary policy sock is given by β Exp. +β Exp.,El. Elasticity i, wic clearly varies by te supply elasticity as long as β Exp.,El. following a contractionary sock is given by β Contr. results are displayed in Table 4. is different from zero. Te + β Contr.,El. Elasticity i. Regression Wile we find tat bot contractionary and expansionary socks ave a greater effect on ouse prices in inelastic areas, tere are nuances to tese results. If we consider a very inelastic market, suc as San Francisco or Miami, te expansionary sock as twice te impact on ouse prices as te contractionary sock, wic is at odds wit te simple demand-supply story. However, considering an area wit very elastic ousing supply, Dayton (OH), te effect of te expansionary sock is smaller tan te contractionary sock. Tis can be due to expansionary socks leading to expectations of substantial future price increases in a market wit a low elasticity of supply, since people may understand tat a ig demand pressure in an inelastic market will lead to iger prices in te future, leading tem to increase demand today in anticipation of future price increases. Tus, 21

22 ouseolds tat are eager to enter te market may increase teir demand today in expectation of increasing prices in te future. At te same time, banks may be willing to extend more credit, since te value of teir ousing portfolio as increased. If price increases lead to expectations of furter price increases, or to a relaxation of credit constraints, tis can ave a strong amplifying effect on demand (Glaeser et al., 2008; Bernanke and Gertler, 1989; Kiyotaki and Moore, 1997; Aoki et al., 2004; Iacoviello, 2005), resulting in large price increases in inelastic markets. On te oter and, an increase in te interest rate may ave a relatively smaller impact on demand. Tese mecanisms will be less relevant in markets wit a ig supply elasticity, since people know tat even in te face of an increase in ouse prices, tere is still plenty of land to build on at a relatively low cost. Figure 8 sow te cumulative response in ouse prices following an expansionary (upper panels) and contractionary (lower panels) monetary policy socks for markets wit an elasticity below average (left panels) and for markets wit an elasticity above average (rigt panels). 22

23 Notes to self: Incomplete and needs to be elaborated. Some ideas: 1. Endogenizing expectations in te teory part, so tat expected ouse price appreciation depends on past price canges Relatively stronger acceleration in inelastic markets 2. Genesove and Mayer (2001) QJE piece on loss aversion may explain smaller drop in demand wen prices are falling. Possible to link to our story? 3. Wat about local differences in credit market institutions? Combination of lax lending standards and low supply elasticities contributing to magnify demand socks? 23

24 Table 4: Effect of contractionary monetary policy sock wit different supply elasticities. =2 =4 =8 =12 =16 Contr. MP sock (0.41) (0.56) (0.89) (1.18) (1.44) Contr. MP sock Elasticity (0.17) (0.21) (0.32) (0.44) (0.52) Exp. MP sock (0.40) (0.78) (1.13) (1.59) (2.07) Exp. MP sock Elasticity (0.18) (0.33) (0.48) (0.67) (0.85) Income per cap (0.07) (0.09) (0.16) (0.21) (0.29) Income per cap. Elasticity (0.03) (0.04) (0.06) (0.07) (0.09) MSA fixed effects Yes Yes Yes Yes Yes Effects of contractionary monetary policy sock for five different MSAs: Dayton (El.=3.64) (0.36) (0.38) (0.60) (0.85) (0.96) Kansas (El.=3.19) (0.30) (0.33) (0.51) (0.72) (0.82) Scranton (El.=1.62) (0.24) (0.31) (0.50) (0.67) (0.82) San Francisco (El.=0.66) (0.32) (0.44) (0.71) (0.94) (1.15) Miami (El.=0.60) (0.33) (0.45) (0.72) (0.96) (1.18) Effects of expansionary monetary policy sock for five different MSAs: Dayton (0.41) (0.58) (0.86) (1.17) (1.49) Kansas (0.35) (0.47) (0.69) (0.93) (1.20) Scranton (0.53) (0.93) (1.29) (1.74) (2.28) San Francisco (0.31) (0.59) (0.85) (1.20) (1.58) Miami (0.32) (0.61) (0.87) (1.24) (1.62) Observations MSAs Witin R Notes: Te dependent variable are te cumulative log canges in te FHFA ouse price index at orizon = 2, 4, 8, 12 and 16. Results are based on estimating equation 8 using a fixed effect estimator and te data set cover a panel of 100 US MSA s countries over te period 1980q1 2007q4. Te specification allows te response in ouse prices to differ depending on te elasticity of supply, as calculated in Saiz (2010), and weter te monetary policy sock is expansionary or contractionary. Standard errors are clustered by MSA and reported in absolute value in parentesis below te point estimates. Te asterisks denote significance levels: * = 10%, ** = 5% and *** = 1%. 24

25 quarters quarters Expansionary, Low Expansionary, Hig quarters quarters Contractionary, Low Contractionary, Hig Figure 8: Asymmetric effect on ouse price from a 1 percentage point monetary policy sock for MSA s wit different ousing supply elasticity, as measured by Saiz (2010). Expansionary, Low refers to te effect of an expansionary monetary policy sock for MSA s wit low supply elasticity, wile Expansionary, Hig refers to te effect of an expansionary monetary policy sock for MSA s wit ig supply elasticity. Likewise, Contractionary, Low refers to te effect of an contractionary monetary policy sock for MSA s wit low supply elasticity, wile Contractionary, Hig refers to te effect of an contractionary monetary policy sock for MSA s wit ig supply elasticity. 5 Conclusion 25

26 References Angrist, J. D., Òscar Jordà, and G. Kuersteiner (2013, August). Semiparametric Estimates of Monetary Policy Effects: String Teory Revisited. NBER Working Papers 19355, National Bureau of Economic Researc, Inc. Anundsen, A. K. and C. Heebøll (2016). Supply restrictions, subprime lending and regional US ousing prices. Journal of Housing Economics 31, Aoki, K., J. Proudman, and G. Vliege (2004). House prices, consumption, and monetary policy: A financial accelerator approac. Journal of Financial Intermediation 13 (4), Auerbac, A. J. and Y. Gorodnicenko (2012). Measuring te output responses to fiscal policy. American Economic Journal: Economic Policy 4 (2), Bernanke, B. S. and M. Gertler (1989). Agency costs, net wort, and business fluctuations. American Economic Review 79 (1), Capozza, D. R., P. H. Hendersott, and C. Mack (2004). An anatomy of price dynamics in illiquid markets: analysis and evidence from local markets. Real Estate Economics 32 (1), Coibion, O. (2012). Are te Effects of Monetary Policy Socks Big or Small? American Economic Journal: Macroeconomics 4 (2), Del Negro, M. and C. Otrok (2007, October). 99 Luftballons: Monetary policy and te ouse price boom across U.S. states. Journal of Monetary Economics 54 (7), Favara, G. and J. Imbs (2015, Marc). Credit Supply and te Price of Housing. American Economic Review 105 (3), Glaeser, E. (2009). Housing supply. Annual Review of Economics 1,

27 Glaeser, E. and J. Gyourko (2005). Urban decline and durable ousing. Journal of Political Economy 113 (2), Glaeser, E., J. Gyourko, and A. Saiz (2008). Housing supply and ousing bubbles. Journal of Urban Economics 64 (2), Glaeser, E. L., J. D. Gottlieb, and J. Gyourko (2012, Marc). Can Ceap Credit Explain te Housing Boom? In Housing and te Financial Crisis, NBER Capters, pp National Bureau of Economic Researc, Inc. Goodart, C. and B. Hofmann (2008). House prices, money, credit, and te macroeconomy. Oxford Review of Economic Policy 24 (1), Green, R. K., S. Malpezzi, and S. K. Mayo (2005). Metropolitan-specific estimates of te price elasticity of supply of ousing, and teir sources. American Economic Review 95 (2), Gyourko, J., A. Saiz, and A. Summers (2008). A new measure of te local regulatory environment for ousing markets. Urban Studies 45 (3), Huang, H. and Y. Tang (2012). Residential land use regulation and te US ousing price cycle between 2000 and Journal of Urban Economics 71 (1), Iacoviello, M. (2005, June). House Prices, Borrowing Constraints, and Monetary Policy in te Business Cycle. American Economic Review 95 (3), Jarocinski, M. and F. Smets (2008). House prices and te stance of monetary policy. Review (Jul), Jordà,. (2005, Marc). Estimation and Inference of Impulse Responses by Local Projections. American Economic Review 95 (1), Jordà, s., M. Scularick, and A. M. Taylor (2015). Betting te ouse. Journal of International Economics 96 (S1), S2 S18. 27

28 Kiyotaki, N. and J. Moore (1997). Credit cycles. Journal of Political Economy 105 (2), Malpezzi, S. (1996). Housing prices, externalities, and regulation in U.S. metropolitan areas. Journal of Housing Researc 7 (2), Malpezzi, S. and S. Wacter (2005). Te role of speculation in real estate cycles. Journal of Real Estate Literature 13 (2), Meen, G. (1990). Te removal of mortgage market constraints and te implications for econometric modelling of UK ouse prices. Oxford Bulletin of Economics and Statistics 52 (1), Meen, G. (2001). Modelling Spatial Housing Markets: Teory, Analysis and Policy. Kluwer Academic Publisers, Boston. Mian, A. and A. Sufi (2009). Te consequences of mortgage credit expansion: Evidence from te U.S. mortgage default crisis. Quarterly Journal of Economics 124 (4), Muellbauer, J. and A. Murpy (1997). Booms and busts in te UK ousing market. Economic Journal 107 (445), Owyang, M. T., V. A. Ramey, and S. Zubairy (2013, May). Are Government Spending Multipliers Greater during Periods of Slack? Evidence from Twentiet-Century Historical Data. American Economic Review 103 (3), Ramey, V. A. (2016, February). Macroeconomic Socks and Teir Propagation. NBER Working Papers 21978, National Bureau of Economic Researc, Inc. Romer, C. D. and D. H. Romer (2004). A new measure of monetary socks: Derivation and implications. American Economic Review 94 (4), Saiz, A. (2010). Te geograpic determinants of ousing supply. Quarterly Journal of Economics 125 (3),

29 Tenreyro, S. and G. Twaites (2016). Pusing On a String: US Monetary Policy is Less Powerful in Recessions. American Economic Journal: Macroeconomics Fortcoming. Tobin, J. (1969). A general equilibrium approac to monetary teory. Journal of Money, Credit and Banking 1 (1), Wieland, J. and M.-J. Yang (2016). Financial dampening. Mimeo, University of California, San Diego. Williams, J. C. (2011, Marc). Monetary Policy and Housing Booms. International Journal of Central Banking 7 (1), Williams, J. C. (2015). Measuring monetary policy s effect on ouse prices. FRBSF Economic Letter. 29

30 Appendix A: 30