Financial Frictions, Distribution Costs, and Current Account Crises Sylvain Leduc1 1 Board 2 Federal
Diego Valderrama2 of Governors
Reserve Bank of San Francisco
2006 Winter Meetings, AEA
Outline Motivation Evidence Related Literature
Outline Motivation Evidence Related Literature Model Overview Details
Outline Motivation Evidence Related Literature Model Overview Details Simulations Solution Method Calibration 4 Experiments Simulation Statistics Importance of the interest rate in business cycles
Outline Motivation Evidence Related Literature Model Overview Details Simulations Solution Method Calibration 4 Experiments Simulation Statistics Importance of the interest rate in business cycles Conclusions
Outline Motivation Evidence Related Literature Model Overview Details Simulations Solution Method Calibration 4 Experiments Simulation Statistics Importance of the interest rate in business cycles Conclusions
Figure: Mexico
Figure: Emerging markets
Empirical regularities to be explained I
Financial imperfections play a large role in EM, at least during recent crises (e.g. Argentina, Brazil, Russia) Interest rates spreads generally increase rapidly in the run-up to the crises and remain high for a period thereafter.
Empirical regularities to be explained I
Financial imperfections play a large role in EM, at least during recent crises (e.g. Argentina, Brazil, Russia) Interest rates spreads generally increase rapidly in the run-up to the crises and remain high for a period thereafter.
I
CA, TB reversals (mostly due import collapse).
Empirical regularities to be explained I
Financial imperfections play a large role in EM, at least during recent crises (e.g. Argentina, Brazil, Russia) Interest rates spreads generally increase rapidly in the run-up to the crises and remain high for a period thereafter.
I
CA, TB reversals (mostly due import collapse).
I
RER depreciations during crises. Burstein, Eichenbaum & Rebelo (2005) find that tradable good prices move more than non-tradable good prices during crises.
Empirical regularities to be explained I
Financial imperfections play a large role in EM, at least during recent crises (e.g. Argentina, Brazil, Russia) Interest rates spreads generally increase rapidly in the run-up to the crises and remain high for a period thereafter.
I
CA, TB reversals (mostly due import collapse).
I
RER depreciations during crises. Burstein, Eichenbaum & Rebelo (2005) find that tradable good prices move more than non-tradable good prices during crises.
I
Ronci (2004) documents a significant fall in volume of trade and trade finance during balance-of-payments crises.
Empirical regularities to be explained I
Financial imperfections play a large role in EM, at least during recent crises (e.g. Argentina, Brazil, Russia) Interest rates spreads generally increase rapidly in the run-up to the crises and remain high for a period thereafter.
I
CA, TB reversals (mostly due import collapse).
I
RER depreciations during crises. Burstein, Eichenbaum & Rebelo (2005) find that tradable good prices move more than non-tradable good prices during crises.
I
Ronci (2004) documents a significant fall in volume of trade and trade finance during balance-of-payments crises.
I
Tornell (2003) argues that one of the main problems in financial crises is the slowdown in the NT sector, which creates bottlenecks in production.
Interest rates in SOE I
Mendoza (AER, 1991) finds that interest rates play a small role in explaining business cycles in a model of a advanced SOE.
I
Oviedo (2004) introduces working capital into model buy finds that standard interest rates shocks provide little additional amplification.
I
Neumeyer & Perri (JME 2004) find that interest rates play a large role in explaining business cycles in EM.
I
Uribe & Yue (JIE 2005) find that once you take into account that interest rates respond exogenously to domestic conditions (e.g. due to increased default probability as in Arellano (2004)) then exogenous interest rate shocks play only a limited role in explaining business cycles.
Trade and distribution costs I
Burstein, Neves & Rebelo (2003) distribution sector and perfect competition to improve the dynamics of the real exchange rate in a model of ERBS.
I
Corsetti & Dedola (2003) two-country model with imperfect competition and pricing to market to derive expression of pass-through as a function of fundamentals.
I
Burstein, Eichenbaum & Rebelo (2002) tightening of borrowing constraints.
I
Trade cost literature, Anderson & van Wincoop (2004), Anderson (2003) find that unmeasured, trade costs important to explain trade patters. Anderson: Insecurity, imperfect contract enforcement are important trade costs.
Outline Motivation Evidence Related Literature Model Overview Details Simulations Solution Method Calibration 4 Experiments Simulation Statistics Importance of the interest rate in business cycles Conclusions
Elements of model
I
Small open economy (SOE) multi-sector model.
Elements of model
I
Small open economy (SOE) multi-sector model.
I
Capacity utilization.
Elements of model
I
Small open economy (SOE) multi-sector model.
I
Capacity utilization.
I
Allow an additional channel for the interest rate to affect output: distribution sector.
Elements of model
I
Small open economy (SOE) multi-sector model.
I
Capacity utilization.
I
Allow an additional channel for the interest rate to affect output: distribution sector.
I
Distribution services needed for domestic consumption of exportables and importables.
Elements of model
I
Small open economy (SOE) multi-sector model.
I
Capacity utilization.
I
Allow an additional channel for the interest rate to affect output: distribution sector.
I
Distribution services needed for domestic consumption of exportables and importables.
I
Working capital constraint on distribution services provides an additional amplification of interest rate shocks.
Households I
Households maximize lifetime utility given by a stationary cardinal utility index. Discount factor is state dependent and induces stationarity in consumption decision in a SOE.
I
Values consumption of nontradables, importables, and exportables.
I
Labor/leisure choice. Greenwood, Herkowitz, and Huffman instantaneous utility that removes wealth effect from labor decision.
I
Households own shares in firms.
I
Households can smooth consumption by buying and selling international bonds, which have an exogenously-determined one-period interest rate.
Goods production
I
Firms make utilization, labor demand decisions to produce tradable and non-tradable goods competitively.
I
Firms make investment decisions. There is a quadratic investment adjustment cost.
I
Traded goods must first be intermediated by distribution firms before they can reach domestic consumers.
I
Firms pay out dividends to households.
I
Firms discount rate taken as given.
I
At the beginning of the period, firms have to set aside portion of wage bill (working capital).
Firms Goods (N,E) producing firms maximize discounted value of dividends:
max
{Li ,I i ,U i }
E0
Πts=0 Mss−1 divi
(1a)
subject to ¯ i Y i − w i Li − I i − K i divi = P t t t t t t
χ 2
Iti (U i )τ − t i τ Kt
2
i
− r ∗ ξ H wLti , (1b)
Yti
i
= exp(ζ )
i Kt+1 = Kti
for i = {N , E}.
1−αi i αi Uti Kti (Lt ) , τ ! i Ut + Iti 1− τ
(1c) (1d)
Firms Goods (N,E) producing firms maximize discounted value of dividends:
max
{Li ,I i ,U i }
E0
Πts=0 Mss−1 divi
(1a)
subject to ¯ i Y i − w i Li − I i − K i divi = P t t t t t t
χ 2
Iti (U i )τ − t i τ Kt
2
i
− r ∗ ξ H wLti , (1b)
Yti
i
= exp(ζ )
i Kt+1 = Kti
for i = {N , E}.
1−αi i αi Uti Kti (Lt ) , τ ! i Ut + Iti 1− τ
(1c) (1d)
Firms Goods (N,E) producing firms maximize discounted value of dividends:
max
{Li ,I i ,U i }
E0
Πts=0 Mss−1 divi
(1a)
subject to ¯ i Y i − w i Li − I i − K i divi = P t t t t t t
χ 2
Iti (U i )τ − t i τ Kt
2
i
− r ∗ ξ H wLti , (1b)
Yti
i
= exp(ζ )
i Kt+1 = Kti
for i = {N , E}.
1−αi i αi Uti Kti (Lt ) , τ ! i Ut + Iti 1− τ
(1c) (1d)
Distribution sector
I
Distribution firms purchase intermediate tradable goods from goods-producing firms and sell them in the domestic market.
I
Must use a certain fraction of produced distribution services to combine with intermediate goods before they can be sold to consumers.
I
Certain fraction of distribution costs are interest rate-sensitive.
Distribution Sector max
{LD ,I D ,U D ,P E ,P M }
E0
Πts=0 Mss−1 divD
(2a)
subject to divD =
X i
¯ i X i − w i Li − I D − K i χ Pti − P t t t t t t 2
i
ηE XtE + ηM XtM
Iti (U i )τ − t i τ Kt
−r ∗ ξ H wLti + r ∗ BtD , 1−αD = exp(ζ D ) UtD KtD (LtD )αD
− BtD = ξBE XtE + ξBM XtM , τ ! UtD D D Kt+1 = Kt 1 − + ItD τ
2
(2b) (2c) (2d) (2e)
Distribution Sector max
{LD ,I D ,U D ,P E ,P M }
E0
Πts=0 Mss−1 divD
(2a)
subject to divD =
X i
¯ i X i − w i Li − I D − K i χ Pti − P t t t t t t 2
i
ηE XtE + ηM XtM
Iti (U i )τ − t i τ Kt
−r ∗ ξ H wLti + r ∗ BtD , 1−αD = exp(ζ D ) UtD KtD (LtD )αD
− BtD = ξBE XtE + ξBM XtM , τ ! UtD D D Kt+1 = Kt 1 − + ItD τ
2
(2b) (2c) (2d) (2e)
Distribution Sector max
{LD ,I D ,U D ,P E ,P M }
E0
Πts=0 Mss−1 divD
(2a)
subject to divD =
X i
¯ i X i − w i Li − I D − K i χ Pti − P t t t t t t 2
i
ηE XtE + ηM XtM
Iti (U i )τ − t i τ Kt
−r ∗ ξ H wLti + r ∗ BtD , 1−αD = exp(ζ D ) UtD KtD (LtD )αD
− BtD = ξBE XtE + ξBM XtM , τ ! UtD D D Kt+1 = Kt 1 − + ItD τ
2
(2b) (2c) (2d) (2e)
Household preferences
max
{CtN ,CtE ,CtM ,Lt ,Bt+1 }
E0
(∞ " X
u(Ct , Lt ) exp −
t−1 X
!#) ¯τ , L ¯τ ) v(C
τ =0
t=0
(3a) where: h
Ct −
i Ltω 1−θ ω
−1
, 1−θ ¯ω ¯t , L ¯ t ) = β ln 1 + C ¯ t − Lt , v(C ω h i1 φ φ φ 1−φ Ct = aT (CtT ) + (1 − aT )1−φ (CtN ) , h i1 ρ ρ ρ 1−ρ . CtT = aT (CtE ) + (1 − aE )1−ρ (CtM ) u(Ct , Lt ) =
(3b) (3c) (3d) (3e)
Households budget constraint
Pt Ct +
X
i qti Kt+1
i
τ Uti + = 1− + τ i X M Kti divit + P t Bt 1 + r ∗0 + wt Lt , (4) M P t Bt+1
i
for i ∈ {N , E, D}.
X
qti Kti
Resource constraints The resource constraints for nontradable goods and tradable goods are: i 2 It (Uti )τ N N N iχ (5) Yt = Ct + It + Kt − 2 Kti τ E Pt
CtE
+
ItE
+
M
M
+ P t CtM + P t E
2 ! Iti (Uti )τ − 2 Kti τ i 2 ! It (Uti )τ M D iχ It Kt − + P t Bt+1 2 Kti τ
χ Kti
M
M
= P t YtE + P t (1 + rt∗ )Bt + rt∗ P t BtD − rt∗ ξLN LtN wt − rt∗ ξH E LtE wt − rt∗ ξLD LtD wt
(6)
Productivity
Productivity for each sector, zti , is assumed to follow an autoregressive process, with mean µz i and correlation κz i : i zti = (1 − κz )µz i + κz i zt−1 + zt ,
(7)
rt represents an innovation to productivity and it is assumed to be white noise.
International Capital Markets
I
One-period international bonds, incomplete markets.
I
Interest rate consists of: international interest rate (Real U.S. 90-day treasury) and country dependent spread.
I
Domestic agents take as given international interest rate. rt∗
∗
= (1 − κr )r + S
Bt + BtD Y
∗ + κr rt−1 + rt ,
S (S(B/Y ) > 0, S 0 (B/Y ) < 0) captures the interest rate premium.
(8)
Outline Motivation Evidence Related Literature Model Overview Details Simulations Solution Method Calibration 4 Experiments Simulation Statistics Importance of the interest rate in business cycles Conclusions
Schmitt-Grohe Uribe algorithm
I
Linearized first order conditions around deterministic steady state for impulse response experiments.
I
Second order approximation to problem with multiple state variables.
Calibration
Parameter σ ω 1/(1 − φ) 1/(1 − ρ) αE αN αD δ χ PM CM /PY PN YN /PY (ηPD )/PE ξLi ξBE , ξBM rr sp
Description Value Coefficient of relative risk aversion 2.6 Labor elasticity of subs. 1.455 Elasticity of subs., T and NT 0.76 Elasticity of subs., X and M 1.5 Labor share of X sector 0.68 Labor share of N sector 0.68 Labor share of D sector 0.68 Depreciation rate 0.025 Investment adjustment cost Trade deficit to GDP +0.01 Consumption import share 0.10 Non-tradable share 0.53 Distribution share 0.50 Working capital 0.50 Share of tradables to pay int. costs 0.50 Real interest rate 0.0% Average Spread 5.0%
Source Ostry Reinhart (1992) Mendoza (1991) Ostry Reinhart (1992)
match σI / σY About right debt-to-GDP ratio Mexico (88–96)
Burstein, Neves, Rebelo Neumeyer and Perri 90-day T-bill - GDP deflator Average EMBI-G spread
Results I
Four experiments: Benchmark: standard model multi-sectoral model. Add distribution sector, sectoral working capital requirements, and borrowing costs for distribution of tradables.
Results I
Four experiments: Benchmark: standard model multi-sectoral model. Add distribution sector, sectoral working capital requirements, and borrowing costs for distribution of tradables.
I
Standard SOE model: multi-sectoral SOE model with capacity utilization.
Results I
Four experiments: Benchmark: standard model multi-sectoral model. Add distribution sector, sectoral working capital requirements, and borrowing costs for distribution of tradables.
I
Standard SOE model: multi-sectoral SOE model with capacity utilization.
I
Standard SOE model with working capital: 50% of wage bill to be financed within period.
Results I
Four experiments: Benchmark: standard model multi-sectoral model. Add distribution sector, sectoral working capital requirements, and borrowing costs for distribution of tradables.
I
Standard SOE model: multi-sectoral SOE model with capacity utilization.
I
Standard SOE model with working capital: 50% of wage bill to be financed within period.
I
Standard SOE model with distribution: 50% of final goods prices reflect distribution services.
Results I
Four experiments: Benchmark: standard model multi-sectoral model. Add distribution sector, sectoral working capital requirements, and borrowing costs for distribution of tradables.
I
Standard SOE model: multi-sectoral SOE model with capacity utilization.
I
Standard SOE model with working capital: 50% of wage bill to be financed within period.
I
Standard SOE model with distribution: 50% of final goods prices reflect distribution services.
I
All effects: distribution, working capital, and distribution sector has interest rate sensitive component.
Standard SOE Model
2: Working capital
3: Distribution
4: Working capital and distribution sector borrowing
Comparing responses
Some intuition
I
Distribution sector is a big intermediate input to consumption. Distribution prices represent about 50% of final prices.
I
Distribution services become a lot more costly due to interest rate increase. Additional negative wealth effect.
Table: Summary statistics: Mexico GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.025 0.024 0.026 0.021 0.021 0.019 0.107
σi /σy 1.000 1.202 1.173 1.265 1.023 1.045 0.921 5.205
σi,y 1.000 0.911 0.977 0.917 -0.753 -0.715 -0.660 -0.503
σt,t−1 0.851 0.866 0.854 0.883 0.900 0.870 0.853 0.894
Table: Simulation summary statistics: Benchmark model GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.030 0.030 0.026 0.016 0.010 0.029 0.003
σi /σy 1.000 1.441 1.425 1.268 0.792 0.500 1.398 0.149
σi,y 1.000 0.638 0.740 0.955 -0.040 0.883 -0.111 -0.062
σt,t−1 0.875 0.875 0.834 0.859 0.834 0.888 0.874 0.833
Table: Summary statistics: Mexico GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.025 0.024 0.026 0.021 0.021 0.019 0.107
σi /σy 1.000 1.202 1.173 1.265 1.023 1.045 0.921 5.205
σi,y 1.000 0.911 0.977 0.917 -0.753 -0.715 -0.660 -0.503
σt,t−1 0.851 0.866 0.854 0.883 0.900 0.870 0.853 0.894
Table: Simulation summary statistics: Working capital GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.032 0.031 0.026 0.018 0.014 0.029 0.003
σi /σy 1.000 1.515 1.502 1.264 0.872 0.675 1.389 0.165
σi,y 1.000 0.593 0.711 0.957 -0.055 0.858 -0.401 -0.073
σt,t−1 0.874 0.874 0.835 0.858 0.837 0.885 0.874 0.836
Table: Summary statistics: Mexico GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.025 0.024 0.026 0.021 0.021 0.019 0.107
σi /σy 1.000 1.202 1.173 1.265 1.023 1.045 0.921 5.205
σi,y 1.000 0.911 0.977 0.917 -0.753 -0.715 -0.660 -0.503
σt,t−1 0.851 0.866 0.854 0.883 0.900 0.870 0.853 0.894
Table: Simulation summary statistics: Distribution GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.046 0.031 0.027 0.015 0.010 0.029 0.006
σi /σy 1.000 2.184 1.467 1.270 0.725 0.468 1.390 0.295
σi,y 1.000 0.234 0.808 0.956 -0.209 0.953 -0.221 -0.216
σt,t−1 0.869 0.856 0.835 0.852 0.833 0.883 0.874 0.833
Table: Summary statistics: Mexico GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.025 0.024 0.026 0.021 0.021 0.019 0.107
σi /σy 1.000 1.202 1.173 1.265 1.023 1.045 0.921 5.205
σi,y 1.000 0.911 0.977 0.917 -0.753 -0.715 -0.660 -0.503
σt,t−1 0.851 0.866 0.854 0.883 0.900 0.870 0.853 0.894
Table: Simulation summary statistics: All effects GDP Tradable output Nontradable output Investment Trade balance Current account Interest rate Real exchange rate
σi 0.021 0.049 0.031 0.027 0.016 0.014 0.029 0.005
σi /σy 1.000 2.312 1.462 1.264 0.770 0.686 1.379 0.253
σi,y 1.000 0.271 0.777 0.955 -0.158 0.933 -0.493 -0.039
σt,t−1 0.870 0.860 0.840 0.854 0.841 0.881 0.874 0.840
Importance of the interest rate in business cycles
If we shut down spread shocks, the volatility of output falls: I
Benchmark case: 2.1% to 2.06%
I
Working capital: 2.1% to 1.9%
I
All effects: 2.1% to 1.8%
Outline Motivation Evidence Related Literature Model Overview Details Simulations Solution Method Calibration 4 Experiments Simulation Statistics Importance of the interest rate in business cycles Conclusions
Concluding remarks
I
Provide a channel to amplify role of interest rates in explaining business cycles.
I
Next step: solve model nonlinearly and estimate unobservable parameters so that model has the best opportunity to reproduce observed behavior (Valderrama 2003).
Additional graphics
Comparing output responses
Comparing trade balance responses
Comparing import responses
Comparing real exchange rate responses
Comparing P N responses
Comparing current account rate responses