Housing and the Business Cycle

Transcription

1 Housing and the Business Cycle Morris Davis and Jonathan Heathcote Winter 2009 Huw Lloyd-Ellis () ECON917 Winter / 21

2 Motivation Need to distinguish between housing and non housing investment,! produced using di erent technologies,! di erent rates of depreciation,! housing yields "home production" services not in National Accounts "Stylized facts" for models with heterogeneous capital goods: (1) comovement between consumption and investment in di erent assets (2) residential investment is 2x as volatile as business investment (3) residential investment leads cycle, business investment lags it Huw Lloyd-Ellis () ECON917 Winter / 21

3 Model Overview Households consume goods and housing services, supply land and labour " Real estate developers combine land and structures to build houses " Two nal goods sector one produces structures; the other C and K " Three intermediate sectors construction, manufacturing and services " Labour, capital and productivity shocks Huw Lloyd-Ellis () ECON917 Winter / 21

4 Main Findings Purely neoclassical model can account for "puzzles" (1) and (2), but not (3) Also matches facts on relative volatility of sub-sectors Implies pro-cyclical house prices, but not volatile enough Why positive comovement and high volatility?,! not due to correlated shocks,! nal goods sectors use all intermediates,! housing requires land, which acts like an adjustment cost,! residential investment is relatively labour intensive,! low depreciation of housing Huw Lloyd-Ellis () ECON917 Winter / 21

5 Population Gross population growth: η > 1 All variables in per capita terms Huw Lloyd-Ellis () ECON917 Winter / 21

6 Intermediate sectors Intermediate rms output: x it = k θ i it (z itn it ) 1 θ i i 2 fb, m, sg,! rent capital at rate r t and hire labour at w t,! output prices p it,! productivity shocks: where ^z t+1 = B^z t + ε t+1 ε t+1 N(0, V) ^z t = [ln z bt, ln z mt, ln z st ] 0 ln z it = ln z it t ln g zi ln z i0 Huw Lloyd-Ellis () ECON917 Winter / 21

7 Final goods sectors Final goods output: y jt = b B j jt mm j jt ss j jt j 2 fc, dg, S j = 1 B j M j,! output prices, p ct = 1 (numeraire) and p dt = relative price of residential investment Huw Lloyd-Ellis () ECON917 Winter / 21

8 Land and real estate Each household sells one unit of land each period Developers combine new residential structures, x d, and new land, x l, to build houses: y ht = x φ φ Structures depreciate at rate δ s lt x 1 dt Total stock of "e ective" housing: ηh t+1 = x 1 dt φ x φ lt + (1 δ s ) 1 φ h t Let 1 δ h = (1 δ s ) 1 φ Huw Lloyd-Ellis () ECON917 Winter / 21

9 Households Optimization problem: max E 0 t=0 β t η t U(c t, h t, n t ) s.t. c t + ηk t+1 + ηp ht h t+1 = (1 τ n )w t n t + [1 (1 τ k ) (r t δ k )] k t +(1 δ h )p ht h t + p lt x lt + ξ t where U(c t, h t, n t ) = 1 1 σ c µ c t h µ h t (1 n t ) 1 µ c µ h 1 σ Huw Lloyd-Ellis () ECON917 Winter / 21

10 First order conditions: where U c (t) = U n (t)(1 τ n )w t U c (t) = βe t [U c (t + 1) (1 (1 τ k ) (r t+1 δ k ))] U c (t)p ht = βe t [U c (t + 1) (1 δ h ) p ht+1 + U h (t + 1)] U c (t) = µ c c 1 t c µ c t h µ h t (1 n t ) 1 µ c µ h 1 σ U n (t) = (1 µ c µ h ) (1 n t ) 1 c µ c t U h (t) = µ h h 1 t c µ c t h µ h t (1 n t ) 1 µ c µ h 1 σ h µ h t (1 n t ) 1 µ c µ h 1 σ Huw Lloyd-Ellis () ECON917 Winter / 21

11 Market Clearing Conditions Final goods and real estate Intermediate goods Factor markets c t + ηk t+1 + g t = y ct + (1 δ k ) k t ηh t+1 = y ht + (1 δ h ) h t x dt = y dt x lt = 1 b ct + b dt = x bt m ct + m dt = x mt s ct + s dt = x st k bt + k mt + k st = k t n bt + n mt + n st = n t Huw Lloyd-Ellis () ECON917 Winter / 21

12 Government budget constraint: ξ t + g t = τ n w t n t + τ k (r t δ k ) k t Huw Lloyd-Ellis () ECON917 Winter / 21

13 Equilibrium prices Factor prices r t = p it θ i k θ i 1 it (z it n it ) 1 θ i i 2 fb, m, sg w t = z it p it (1 θ i )k θ i it (z itn it ) θ i Prices of intermediates y ct p p bt = B c = dt y dt B d b ct b dt y ct p p mt = M c = dt y dt M d m ct m dt p st = S c y ct s ct Prices of structures and land = S d p dt y dt s dt p dt = (1 φ) p hty ht x dt p lt = φ p hty ht x lt Huw Lloyd-Ellis () ECON917 Winter / 21

14 Implication for House prices Relative price of residential investment can be written as ln p dt = (B c B d ) (1 θ b ) ln z bt + (M c M d ) (1 θ m ) ln z mt + (S c S d ) (1 θ s ) ln z st + other terms,! a positive shock in sector i will reduce p dt if residential investment is relatively intensive in input i ) implications for comovement Price of new housing ln p ht = ln(1 φ) + φ ln y dt + ln p dt Huw Lloyd-Ellis () ECON917 Winter / 21

15 Mapping between model and NIPA In NIPA private consumption includes imputed value for rents from owner occupied housing: where PCE t = c t + q t h t q t = U h(c t, h t, n t ) U c (c t, h t, n t ) In NIPA, raw land is not part of GDP ) should only include value of residential investment, not of new houses: GDP t = y ct + p dt y dt + q t h t Real private consumption and GDP de ned using balanced growth prices ) does not capture short-run price movements Huw Lloyd-Ellis () ECON917 Winter / 21

16 Balanced Growth Path Although each sector has di erent growth rates, a BGP exists due to Cobb-Douglas assumptions All variables are made stationary by dividing by gross growth rate Model is solved using Klein (2000) ˆx t = x t g t x Huw Lloyd-Ellis () ECON917 Winter / 21

17 Table 1: Growth Rates on Balanced Growth Path (growth rates gross, variables per-capita) n b, n m, n s, n, r 1 1 [ ] Bcθ b M cθ m Scθ s k b, k m, k s, k, c, i k, g, y c, w Bc ( 1 θ b ) M c ( 1 θ m ) Sc ( 1 θ s ) g g g g b c, b h, x b k = 1 zb zm zs g b = g g θ b 1 θ b k zb m c, m h, x m s c, s h, x s x d g d g m s = g = g g = g m k g θ θ g 1 m zm θ s 1 θ s k zs Bh b g M h m g Sh s x 1 l g l = η y h, h g h = g g φ 1 φ l d p h y h, p d x d, p l x l, p b x b, p m x m, p s x s g k Table 2: Tax Rates, Depreciation Rates, Adjustment Costs, Preference Parameters Davis Heathcote Grenwood Hercowitz (GH) Tax rate on capital income: τ k Tax rate on labor income: τ n Govt. cons. to GDP 0.179* Transfers to GDP 0.076* Depreciation rate for capital: δ k * Depreciation rate for res. structures: δ s * Land s share in new housing: φ Population growth rate: η * 0.0 Discount factor: β Risk aversion: σ 2.00* 1.00 Consumption s share in utility: µ c Housing s share in utility: µ h Leisure s share in utility: 1-µ c -µ h Starred parameter vales are chosen independently of the model.

18 Calibration Period is one year µ c and µ h chosen so that ˆn = 0.3 and value of stock of residential structures = GDP τ k and τ n chosen so that non-residential capital stock = 1.5 x annual output and ξ/gdp = Shock processes estimated as VAR,! little evidence of spillovers weak correlation of shocks,! shocks to construction and manufacturing much more volatile input shares based on input output tables Huw Lloyd-Ellis () ECON917 Winter / 21

19 Table 3: Production Technologies Con. Man. Ser. GH Input shares in cons/inv production B c, M c, S c Input shares in res. structures B d, M d, S d Capital s share by sector θ b, θ m, θ s Trend productivity growth (%) g zb, g zm, g zs Autocorrelation coefficient see table 4 ρ = 1.0 Std. dev. innovations to logged productivity see table Table 4: Estimation of Exogenous Shock Process System estimated: ~ z ~ t+ 1 = Bz t + ε t+ 1 log ~ zbt ε where zt = log ~ bt ~ zmt, ε log ~ t = ε mt and ε t ~ N(0, V ). 5 z st ε st Autoregressive coefficients in matrix B (Seemingly unrelated regression estimation method: standard errors in parentheses) log ~ z b,t + 1 log ~ z m,t + 1 log ~ z s,t + 1 log ~ z bt (0.089) log ~ z mt (0.083) log ~ z st (0.098) (0.078) (0.073) (0.087) (0.038) (0.036) (0.042) R Correlations of innovations ε b ε m ε s ε b ε b ε m ε m ε s 1 ε s Standard deviation of innovations 5 All variables are linearly detrended prior to estimating this system.

20 Main Results Steady state implications look "reasonable" Second moments:,! accounts well for high relative volatility of residential investment,! yields comovement between investment sectors,! relative volatilities of sub-sectors is correct,! house price volatility is too low,! housing investment does not lead cycle Huw Lloyd-Ellis () ECON917 Winter / 21

21 Table 5: Decomposition of Final Expenditure into Final Sales From Industries (%) (based on 1992 IO-Use Table) PCE BFI + RESI RESI 6 BFI G 7 Construction Manufacturing Services Table 6: Decomposition of Final Expenditure into Value Added by Industry (%) PCE BFI + RESI RESI BFI PCE + BFI + GOVI 8 Construction Manufacturing Services Table 7: Properties of Steady State: Ratios to GDP % Data ( ) Model Capital stock (K) Residential structures stock (P d x S) Private consumption (PCE) Government consumption (G) Non-residential inv (non-resi) Residential inv (P d x RESI) Construction (P b x Y b ) Manufacturing (P m x Y m ) Services (P s x Y s ) Real after tax interest rate (%) We attribute all $225.5 billion of residential investment in 1992 to sales from the construction industry, since the first I/O use table does not have a residential investment column. We defend this choice in the data appendix. 7 G is government expenditure, which includes government consumption and government investment expenditures. 8 GOVI is government investment. We assume that the value-added composition of government investment by intermediate industry is the same as business fixed investment. 9 The shares of construction, manufacturing and services do not add to exactly one, since the product approach to computing GDP does not give exactly the same answer as the expenditure approach. In both model and data, imputed rental income from owner-occupied housing is attributed to the service sector.

22 Table 8: Business Cycle Properties 10 Data ( ) Model % Standard Deviations GDP (rel. to GDP) PCE Labor (N) Non-RESI RESI House prices (P h ) 1.37 ( ) 0.40 Construction output (Y b ) Manufacturing output (Y m ) Services output (Y s ) Construction hours (N b ) Manufacturing hours (N m ) Services hours (N s ) Correlations PCE, GDP P h, GDP 0.65 ( ) 0.65 PCE, non-resi PCE, RESI non-resi, RESI P h, RESI 0.34 ( ) N b, N m N b, N s N m, N s Lead-lag correlations i = 1 i = 0 i = -1 i = 1 i = 0 i = -1 non-resi t-i, GDP t RESI t-i, GDP t non-resi t-i, RESI t Statistics are averages over 500 simulations, each of length 54 periods, the length of our data sample. Prior to computing statistics all variables are (i) transformed from the stationary representation used in the solution procedure back into non-stationary representation incorporating trend growth, (ii) logged, and (iii) Hodrick-Prescott filtered with the smoothing parameter, λ, set to 100.

23 Understanding the results Model GH one sector RBC model with some capital entering (log) utility function Model A σ = 2 and non-permanent shocks Model B adding land Model C sector speci c shocks Model D a distinct technology for residential structures Model E sector speci c capital shares Model F asset speci c depreciation rates (benchmark) Huw Lloyd-Ellis () ECON917 Winter / 21

24 Table 9: Alternative Parameterizations Model Description Selected parameter values GH Greenwood and Hercowitz see tables 2 and 3 A One sector model, housing in utility (re-parameterized GH) σ = 2, ρ = 0.85, σ(ε) = δ k = δ s = θ b = θ m = θ s = 0.25 B d, = B c, M d, = M c, S d, = S c B A + land φ = C B + sector-specific shocks see table 4 D C + two final goods technologies see table 3 E D + sector-specific capital shares θ b = 0.132, θ m = 0.309, θ s = F (Benchmark) E + different depreciation rates δ s = Table 10: Alternative Parameterizations: Business Cycle Properties Data GH A B C D E F GDP (% std dev) Std. dev. relative to GDP PCE N Non-RESI RESI Y b Y m Y s P h Correlations Non-RESI, RESI P h, RESI Lead-lag pattern: corr(x t-1, GDP t ) corr(x t+1, GDP t) x = RESI x = Non-RESI

25 Comparison to US Historical Time Series Model matches trend growth in hours, consumption and output Residential investment exhibits relatively slow growth (due to low prod. growth in construction) Does not match growth of non-residential investment or manufacturing Model does reasonably well at business cycle frequencies Does not account well for house price dynamics Huw Lloyd-Ellis () ECON917 Winter / 21

26

27

28

29 Supply vs. Demand Shocks Traditional view based on regressions of residential investment on house prices (controlling for other factors),! positive coe cient ) demand shocks more important than supply shocks Regression using simulated data from model yields positive coe cient, despite no demand shocks,! due to omitted variable bias Huw Lloyd-Ellis () ECON917 Winter / 21