Learning About Commodity Cycles and Saving-Investment Dynamics in a Commodity-Exporting Economy

Transcription

1 Learning About Commodity Cycles and Saving-Investment Dynamics in a Commodity-Exporting Economy Jorge Fornero Markus Kirchner Central Bank of Chile, Macroeconomic Analysis Division Fifth BIS CCA Research Conference on Challenges from changing international financial conditions Bogota, Colombia, May 214 The views and conclusions presented are exclusively those of the authors and do not necessarily reflect the position of the Central Bank of Chile or its Board members. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 1 / 2

2 Motivation: Recent CA dynamics in commodity-exporting countries Commodity prices have surged over the past decade, and most commodity exporters have experienced record terms of trade. Despite very favorable terms of trade, many commodity exporters have accumulated sizable current account (CA) deficits. Some have even experienced a CA reversal over time: In Chile, the CA balance moved from a surplus of 4.6% of GDP in 26 to a deficit of -3.4% in A similar process occurred e.g. in Brazil, Canada and Peru. CA deficits have become an important policy concern: Might reflect macroeconomic imbalances such as excessive domestic absorption and over-borrowing. Pose risk of a painful adjustment under sudden stops of capital flows. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 2 / 2

3 Copper price boom and external savings balance in Chile USD per metric pound, % of nominal GDP Copper price (left axis) Current account (left axis) % of nominal GDP -3 National savings (right axis) National investment (right axis) Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 3 / 2

4 Investment boom, commodity cycle and expectations Other stylized facts: Successive upward revisions of long-run copper price forecasts from 26-7 onwards: By professional forecasters. By the panel of experts that counsels on the parameters of Chile s fiscal rule (including a long-run reference price of copper). Rise in national investment after 27 was mainly driven by mining: Investment in mining grew from 2.5% of GDP in to almost 5% in Other investment also increased but less. FDI in mining explained more than half of total FDI in 28-12, and roughly tripled compared to historical volumes. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 4 / 2

5 Evolution of professional copper price forecasts 5 45 Effective spot price of copper Forecasts in October of each year 4 US dollar cents per metric pound Note: Forecasts by CRU Group. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 5 / 2

6 Evolution of the government s reference price of copper 45 4 Reference price of copper Effective real price 35 US dollar cents per metric pound Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 6 / 2

7 Investment in mining vs other sectors 25 Gross fixed capital formation in mining Gross fixed capital formation in other sectors % of nominal GDP Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 7 / 2

8 Objective of this paper Possible mechanism: Agents gradually adjusted their perceptions on the duration of the commodity cycle, which affected savings and investment especially in the commodity sector. Goal: Analyze the role of this mechanism for CA dynamics in a commodity-exporting economy. What do we do? Incorporate two novel elements in an NK-SOE model: Endogenous commodity production with capital and time to build. Imperfect information and learning by agents on the persistence of commodity price shocks. Estimate the model with Chilean data from 21 to 212, following a Bayesian approach. Use the model to analyze the role of learning: To drive the response of macro variables to commodity price shocks. To understand the observed gradual CA adjustment in Chile. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 8 / 2

9 Imperfect information: A simple unobserved components model for the copper price with persistent and transitory shocks Assume that the international price (real terms, log deviations from long-run mean) satisfies: p S,t = a t +b t, t =,1,2,... The unobservable shock a t captures transitory noise : a t NID(,σ 2 a). While b t is an unobserved state variable that measures persistent, fundamental cycles: b t = ρb t 1 +u t, ρ [,1), u t NID(,σ 2 u). Data: Price of refined copper (London Metals Exchange, deflated by trade-adjusted external price index); 196Q1-212Q4. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 9 / 2

10 Learning behavior by agents and estimation of parameters Following Erceg and Levin (23), agents use the Kalman filter (KF) to obtain the optimal linear inference of the state: ˆb t = E[b t p S,t,...,p S, ] = ρˆb t 1 +Kρ 1 (p S,t ρˆb t 1 ). K t is the Kalman gain with Σ t = var[b t ps,t,...,p S, ]. For t, steady state with Σ t = Σ and K t = K. Agents learn at a constant rate, adjusting inference ˆb t and forecasts ˆp S,t+h = ρˆb h t according to past prediction errors. Parameters σ a and σ u not separately identified. Estimation (ρ < 1): 1 Fix K =.15 (Erceg and Levin, 23; Céspedes and Soto, 27). 2 Obtain κ = σ u /σ a =.17 from KF equations that yield K and Σ. 3 Rewrite the model as p S,t = b t +(σ u /κ)ξ t, ξ t NID(,1), b t = ρb t 1 +σ u ζ t, ζ t NID(,1). 4 Obtain estimates by ML: ρ =.979, σ u =.375, σ a =.232. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 1 / 2

11 Copper price decomposition from the KF 1 8 Filtered transitory component Filtered persistent component Effective real copper price % from steady state (average 21Q3-212Q4) Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 11 / 2

12 An NK model of a commodity-exporting small open economy Based on Medina and Soto (27) model used for policy analysis at the Central Bank of Chile. 1 Basic structure similar to Smets and Wouters (23, 27), Christiano et al. (25, CEE), Adolfson et al. (27) models. Key extensions: 1 Endogenous commodity production using capital specific to the commodity sector and a fixed factor (land). 2 The government owns a share χ of total assets, the rest is FDI. Tax on profits of foreign investors. 1 Main elements of the model: Consumption of home goods, imported goods, oil and food; Staggered price-setting à la Calvo with indexation both for domestic producers and importers (i.e. delayed pass-through); Sticky wages à la Calvo with indexation; Labor-augmenting productivity growth; Habits in consumption; Investment adjustment costs; Non-Ricardian households; Taylor rule (smoothing, inflation and GDP growth); Elastic country premium; Commodity sector (endowment, exogenous international price); Structural balance fiscal rule for government spending (consumption, complete home bias). Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 12 / 2

13 Capital accumulation in the commodity sector (sector S) Production Y S,t = F(K S,t 1,...). Slow accumulation of capital: Convex costs of initiating investment projects (CEE, 25; Uribe and Yue, 26). Time to build (Kydland and Prescott, 1982; Uribe and Yue, 26). It takes n 1 periods for investment projects to turn productive: K S,t = (1 δ S )K S,t 1 +[1 Φ S (X S,t n+1 /X S,t n )]X S,t n+1, where X S,t n+1 are investment projects started t n+1 periods ago. Effective investment expenses are I S,t = n 1 j= ϕ jx S,t j, where ϕ j is fraction of projects initiated in t j and financed in t. I S,t is a CES bundle of domestic goods (e.g. construction) and imported goods (e.g. machinery). Generates spillover effects on non-commodity production and the trade balance. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 13 / 2

14 Cash flow maximizing capital-investment choice FOC, general case: K S,t : Q S,t = E t P C,t Λ t,t+1 X S,t : ϕ P IS,t P C,t +ϕ 1 E t Q S,t+1 P C,t+1 (1 δ S ) + P S,t+1A S F S K (T S t+1,k S,t ) P C,t+1 { Λ t,t+1 P IS,t+1 P C,t+1 + +ϕ n 1 E t { Λ t,t+n 1 P IS,t+n 1 = E t Λ t,t+n 1 Q S,t+n 1 P C,t+n 1 { P IS,t+2 }+ϕ 2 E t Λ t,t+2 1 Φ S } P C,t+n 1 ( ) XS,t ( X S,t 1 ) Φ XS,t XS,t S X S,t 1 X S,t 1 ( )( ) 2 Q +Λ S,t+n t,t+n P C,t+n Φ XS,t+1 XS,t+1 S X S,t X S,t P C,t+2 Higher forecasted commodity price (P S,t+1 ) stimulates investment. Due to TTB, future capital prices (Q S,t+n 1, Q S,t+n ) matter: Only (perceived) persistent commodity price increases affect investment. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 14 / 2 }

15 Government, foreign investors and CA balance A fraction χ of the cash flow from sector S goes to the government, plus taxes on profits of foreign investors. Government spending follows a structural balance rule linking it, inter alia, to a long-run reference price P S,t, i.e. the forecast of P S,t averaged over a 1 years horizon. Hence, (perceived) transitory price increases are mostly saved while a higher long-run price allows more spending. The CA balance then equals the change in the international investment position of the economy: CAY t = ε tbt 1 P Y,t Y t (1+it ε tbt 1 1 )Θ t P Y,t Y t (1+it 1 )Θ t 1 }{{} Change in portfolio investment position (1 χ) Q S,t(K S,t K S,t 1 ). P Y,t Y t }{{} Change in FDI position Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 15 / 2

16 Results: IRFs to persistent and transitory commodity price shocks Obs. commodity price Filtered persistent comp. Filtered transitory comp Δ% from st. state GDP Private consumption Investment Δ% from st. state Inflation Real exch. rate Current account / GDP Δ% from st. state Quarters after shock Quarters after shock Quarters after shock Persistent shock, imperfect information Persistent sh., full info. Transitory sh., full info. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 16 / 2

17 Components of the CA (persistent shock, imperfect information) Inv. sector S Inv. non-s Private savings Gov. savings CA / GDP % from steady state Quarters after shock Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 17 / 2

18 Closing the commodity investment channel Total GDP Non-S GDP GDP sector S Δ% from st. state Total investment Non-S investment Inv. sector S Δ% from st. state Non-S Tobin's q Tobin's q sector S Current account / GDP Δ% from st. state Quarters after shock Quarters after shock Quarters after shock Endogenous commodity production Exogenous commodity production Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 18 / 2

19 Historical decomposition of Chile s CA balance 8 C. Current account balance (% of GDP) 6 Δ% from steady state Persistent copper price shock Foreign interest rate shocks Smoothed variable Transitory copper price shock Other shocks and init. values Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 19 / 2

20 Conclusions The interaction of commodity-specific investment and learning by agents is crucial to explain the gradual CA deterioration: Higher savings in the short run, as agents fear that the commodity boom might be temporary. Lower savings and higher investment afterwards, as agents learn about the actual persistence of the shock. Policy implications: A gradual CA deterioration is not necessarily a sign of emerging imbalances: Product of investment in the commodity sector? CA deficits due to FDI seem less worrisome than others, e.g. due to large portfolio inflows. Anyhow, limited scope for monetary policy to affect investment and FDI in the commodity sector (not shown, see paper). Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 2 / 2

21 Appendix Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 21 / 2

22 Production and firms problem in sector S Technology is Cobb-Douglas with decreasing returns to capital: Y S,t = A S,t T η S t K 1 η S S,t 1, where A S,t captures technological shocks specific to sector S (e.g. variations in mineral content of land), while T t is growth trend. Define gross profits as: where κ S are fixed costs. Π S,t = P S,t Y S,t P C,t T t κ S, The nominal flow of investment is P IS,tI S,t, and the firm is assumed to maximize its real cash flow CF S,t = Π S,t P IS,tI S,t : maxe t i= Λ t,t+i CF S,t+i P C,t+i. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 22 / 2

23 Government balance and fiscal rule A fraction χ of the cash flow from sector S goes to the government, plus taxes τ S on profits of foreign investors. Government balance: ε t B G,t (1+i t )Θ t +P G,t G t = ε t B G,t 1 +τ tp Y,t Y t +χcf S,t +τ S (1 χ)(π S,t δ S Q S,t K S,t 1 ), where τ t are lump-sum taxes on HHs (a fraction of nominal GDP). Government spending follows the structural balance rule: ( ) 1 εtb P G,t G t 1 = (1+i G,t 1 τtpy,tȳt t 1 )Θ t 1 P Y,t Y t + P Y,t Y t +χ CF S,t P Y,t Y t. P Y,t Y t +τ S (1 χ) Π S,t δ S Q S,t K S,t 1 P Y,t Y t VCt P Y,t Y t target P Y Y The term VC t = [χ+τ S (1 χ)]y S,t ε t (PS,t P S,t ) is a cyclical adjustment that depends on the difference of PS,t and the long-run reference price P S,t. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 23 / 2

24 Estimation of the model and computation of IRFs Estimation strategy: Given exogeneity of the copper price for Chile, observe ˆb t. The state space representation to compute the likelihood is standard: Y t = Hx t +v t, v t NID(,Σ v ), x t = Dâ t +Eˆb t +Fx t 1 +Gǫ t, ǫ t NID(,Σ ǫ ). With priors for a subset of parameters (leaving others calibrated), we estimate the model with Bayesian techniques. Computation of IRFs to commodity price shocks: 1 Assume a persistent or transitory shock to the actual price p S,t, and calculate the inferred persistent component ˆb t using the KF recursion. 2 With ˆb t and p S,t, compute â t = p S,t ˆb t. 3 Given ˆb t and â t, simulate the response of the economy. Fornero & Kirchner Learning About Commodity Cycles and S-I Dynamics in a Commodity-Exporting Economy 24 / 2