Commodity Price Beliefs, Financial Frictions and Business Cycles

Transcription

1 Commodity Price Beliefs, Financial Frictions and Business Cycles Jesús Bejarano Franz Hamann Enrique G. Mendoza 1 Diego Rodríguez Preliminary Work Closing Conference - BIS CCA Research Network on The commodity cycle: macroeconomic and financial stability implications Mexico City, August University of Pennsylvania, NBER and PIER. Remaining authors are at Banco de la República.

2 Commodity cycle challenges macro and financial stability I Commodity price volatility has affected commodity-exporters macro performance, through now well-known changes in incentives to borrow/lend in presence of financial frictions 2 2 See Mendoza (1991;95), Uribe & Yue (2006), Bianchi, Boz & Mendoza (2012), Bianchi & Mendoza (2015)]

3 Commodity cycle challenges macro and financial stability I Commodity price volatility has affected commodity-exporters macro performance, through now well-known changes in incentives to borrow/lend in presence of financial frictions 2 I This kind of risk to oil exporting economies is usually uninsured in international financial markets and the macro and financial adjustment works through the real exchange rate (RER) and net foreign financial assets (NFA) 2 See Mendoza (1991;95), Uribe & Yue (2006), Bianchi, Boz & Mendoza (2012), Bianchi & Mendoza (2015)]

4 Commodity cycle challenges macro and financial stability I Commodity price volatility has affected commodity-exporters macro performance, through now well-known changes in incentives to borrow/lend in presence of financial frictions 2 I This kind of risk to oil exporting economies is usually uninsured in international financial markets and the macro and financial adjustment works through the real exchange rate (RER) and net foreign financial assets (NFA) I Furthermore, with financial frictions (collateral constraints) there is a pecuniary externality: agents do not internalize when they borrow in good times that high leverage causes collapse in collateral values and credit crunch in bad times (Fisherian deflation) 2 See Mendoza (1991;95), Uribe & Yue (2006), Bianchi, Boz & Mendoza (2012), Bianchi & Mendoza (2015)]

5 Uncertainty reflected in spot prices as well as in futures Q3 WTI Oil Price (USD per Barrel) Q M1 00-M7 01-M1 01-M7 02-M1 02-M7 03-M1 03-M7 04-M1 04-M7 05-M1 05-M7 06-M1 06-M7 07-M1 07-M7 08-M1 08-M7 09-M1 09-M7 10-M1 10-M7 11-M1 11-M7 12-M1 12-M7 13-M1 13-M7 14-M1 14-M7 15-M1 15-M7 16-M1 16-M7 Month Spot Jul-14 Aug-14 Sep-14 Jan-15 Feb-15 Mar-15

6 There is uncertainty about commodity price fundamentals I Under-appreciating the true process of commodity prices is natural because of uncertainty about the duration of the high/low price regimes, as well as the lack of knowledge about the true factors behind commodity price fundamentals.

7 There is uncertainty about commodity price fundamentals I Under-appreciating the true process of commodity prices is natural because of uncertainty about the duration of the high/low price regimes, as well as the lack of knowledge about the true factors behind commodity price fundamentals. I Natural resources are known to be affected by political instability, changes in their market structure, structural changes in technology to exploit them, shifts in global demand, etc.

8 There is uncertainty about commodity price fundamentals I Under-appreciating the true process of commodity prices is natural because of uncertainty about the duration of the high/low price regimes, as well as the lack of knowledge about the true factors behind commodity price fundamentals. I Natural resources are known to be affected by political instability, changes in their market structure, structural changes in technology to exploit them, shifts in global demand, etc. I As a result, discovering the true process of commodity prices is an ongoing business as the academic debate also illustrates: Hamilton (2003), Rebucci and Spatafora (2006), Hamilton (2008), Kilian (2009) and Baumeister & Kilian (2015).

9 Commodity price beliefs interact with financial frictions I When there is uncertainty about commodity price fundamentals, optimistic beliefs or good news about commodity prices strengthen incentives to borrow today and increase expected future borrowing capacity 3 but also... 3 Cogley and Sargent (2008), Boz (2009) and Boz, Daude and Durdu (2009)

10 Commodity price beliefs interact with financial frictions I When there is uncertainty about commodity price fundamentals, optimistic beliefs or good news about commodity prices strengthen incentives to borrow today and increase expected future borrowing capacity 3 but also... I optimistic beliefs change the incentives to exploit natural resources (real assets) affecting reserves and extraction of the commodity 3 Cogley and Sargent (2008), Boz (2009) and Boz, Daude and Durdu (2009)

11 Commodity price beliefs interact with financial frictions I When there is uncertainty about commodity price fundamentals, optimistic beliefs or good news about commodity prices strengthen incentives to borrow today and increase expected future borrowing capacity 3 but also... I optimistic beliefs change the incentives to exploit natural resources (real assets) affecting reserves and extraction of the commodity I... and if followed by disappointing price outcomes, the probability of a financial crisis rises because of higher leverage and the endogenous cutback in commodity extraction. 3 Cogley and Sargent (2008), Boz (2009) and Boz, Daude and Durdu (2009)

12 Commodity price beliefs interact with financial frictions I When there is uncertainty about commodity price fundamentals, optimistic beliefs or good news about commodity prices strengthen incentives to borrow today and increase expected future borrowing capacity 3 but also... I optimistic beliefs change the incentives to exploit natural resources (real assets) affecting reserves and extraction of the commodity I... and if followed by disappointing price outcomes, the probability of a financial crisis rises because of higher leverage and the endogenous cutback in commodity extraction. I In this paper, discrepancies between initial expectations about prices and actual and posterior expected prices can be an important source of macro and financial instability because it affects non-trivially both households and firms. 3 Cogley and Sargent (2008), Boz (2009) and Boz, Daude and Durdu (2009)

13 Amodelofacommodityexportingeconomy... I Time, t = 0, 1, 2,...,economywithy T, y N and a stock of commodity s > 0, all in fixed supply. Every t there is s t 2 [0, s] extracts x t 2 [0, s t ] and discovers d 0: s t+1 = s t x t + d.

14 Amodelofacommodityexportingeconomy... I Time, t = 0, 1, 2,...,economywithy T, y N and a stock of commodity s > 0, all in fixed supply. Every t there is s t 2 [0, s] extracts x t 2 [0, s t ] and discovers d 0: s t+1 = s t x t + d. I Commodity price p t has a true TPM, Q (p t+1, p t ), unknown to agents. The value of a competitive firm with perfect access to financial markets is: v(s t, p t )= max x2[0,s t] n p t x t e(s t, x t )+R 1 E B t [v (s t+1, p t+1 )] Et B is conditional on the agents beliefs with the information available up to t. Rational expectations: E t. o

15 Amodelofacommodityexportingeconomy... I Time, t = 0, 1, 2,...,economywithy T, y N and a stock of commodity s > 0, all in fixed supply. Every t there is s t 2 [0, s] extracts x t 2 [0, s t ] and discovers d 0: s t+1 = s t x t + d. I Commodity price p t has a true TPM, Q (p t+1, p t ), unknown to agents. The value of a competitive firm with perfect access to financial markets is: v(s t, p t )= max x2[0,s t] n p t x t e(s t, x t )+R 1 E B t [v (s t+1, p t+1 )] Et B is conditional on the agents beliefs with the information available up to t. Rational expectations: E t. I Inter-temporal optimality condition: p t e x (s t, x t )= EB t [p t+1 e x (s t+1, x t+1 ) e s (s t+1, x t+1 )]. R o

16 ... with liability dollarization Let x be the firm s optimal extraction. HH maximize: " X 1 # c t = subject to apple a ct T E B 0 µ +(1 t=0 t c1 t 1 a) c N t µ 1 µ, a > 0,µ 1 c T t +p N t c N t = y T + (p t, s t )+ E e(s t, x (p t, s t ))+p N t y N b t+1 +Rb t with p t x t e (s t, x t ) and S p t x t E e(s t, x t) t and b t+1 y T + (p t, s t )+ E e(s t, x (p t, s t )) + p N t y N.

17 Rational expectations perfect information equilibrium The recursive representation of REPI equilibrium conditions are: apple 1 a c p N T = a y h N i c = RE t c µ b 0 = y T + S p x (p, s) c T + Rb b 0 y T + S p x (p, s)+p N y N

18 Equilibrium under Bayesian learning about commodity prices I Follow Boz and Mendoza (2009) two-stage solution strategy: 1. Bayesian learning. Take a history of price realizations observed up to date t, p t =(p t, p t 1, p t 2,..., p 1 ) and generate a sequence of posterior density functions f Q B p t T t=1 over T periods. Each f is a probability distribution over possible Q B s. The date t = 0 priors depend on the assumed amount of agents prior knowledge.

19 Equilibrium under Bayesian learning about commodity prices I Follow Boz and Mendoza (2009) two-stage solution strategy: 1. Bayesian learning. Take a history of price realizations observed up to date t, p t =(p t, p t 1, p t 2,..., p 1 ) and generate a sequence of posterior density functions f Q B p t T t=1 over T periods. Each f is a probability distribution over possible Q B s. The date t = 0 priors depend on the assumed amount of agents prior knowledge. 2. The problem can be divided into a sequence of AU optimization problems (AUOP) for t = 1, 2,...,T,each conditional on E t q B hh and Et q B ll, where the time indexes identify the date of the beliefs that match the corresponding AUOP. So, we find a sequence of equilibrium policy functions for {x t } T t=1 and {b0 t} T t=1,oneforeachsetofbeliefsateach date t = 1, 2,...,T.

20 Anticipated utility optimal problem The solution to AUOP at date t is given by x t (s, p) which solves n v t (s, p) =max [px e(s, x)] + R 1 E B t vt s x + d, p 0 o, x2x the policies bt 0 (b, s, p), c t (b, s, p), ct T (b, s, p), ct N (b, s, p), t (b, s, p) and a pricing function pt N (b, s, p) that satisfy HH optimality as well as the MCC for T and NT sectors: apple 1 a c pt N T 1+µ (b, s, p) = t (b, s, p) a y N c t (b, s, p) = RE B t ct+1 (b, s, p) + t (b, s, p) bt 0 (b, s, p) = y T + S p x t (s, p) ct T (b, s, p)+rb bt 0 (b, s, p) apple y T + S p x t (s, p)+pt N (b, s, p) y N

21 Initial calibration: oil prices and the Colombian economy I True price process: a hidden Markov model p t = p (I t )+ t where I t is an indicator variable that records whether oil prices are high or low and t is an identically and independently distributed normal random variable with mean 0 and variance 2.

22 Initial calibration: oil prices and the Colombian economy I True price process: a hidden Markov model p t = p (I t )+ t where I t is an indicator variable that records whether oil prices are high or low and t is an identically and independently distributed normal random variable with mean 0 and variance 2. I We apply Hamilton (1989) Markov switching estimator on quarterly real oil prices covering the period 1970:1 to 2014:2. As a proxy for real oil prices we take the BRENT crude oil price in nominal US dollars deflated by the United States Consumer Price Index. The base year for the US CPI is 1983.

23 Initial calibration: oil prices and the Colombian economy I True price process: a hidden Markov model p t = p (I t )+ t where I t is an indicator variable that records whether oil prices are high or low and t is an identically and independently distributed normal random variable with mean 0 and variance 2. I We apply Hamilton (1989) Markov switching estimator on quarterly real oil prices covering the period 1970:1 to 2014:2. As a proxy for real oil prices we take the BRENT crude oil price in nominal US dollars deflated by the United States Consumer Price Index. The base year for the US CPI is I Take the model ergodic moments under REPI to match both aggregate and sectoral (oil) Colombian data.

24 Debt distribution from low to high commodity prices In t = 1, p t = p h and then for t = 2,...,7 p t = p l RE

25 Price collapse after a long period of high prices I Experiment: we date the start of the high commodity price regime in the 2009:4 and its end on 2014:3. The low oil price regime goes from 2014:4 to 2016:1. We assume that agents have uninformative initial priors and experience the high-oil price regime during five years, followed by a year and a half of low prices.

26 Price collapse after a long period of high prices I Experiment: we date the start of the high commodity price regime in the 2009:4 and its end on 2014:3. The low oil price regime goes from 2014:4 to 2016:1. We assume that agents have uninformative initial priors and experience the high-oil price regime during five years, followed by a year and a half of low prices. I Case 1: the sequence of revenues is {p t x t } T t=1 with {x t = d} T t=1.thiscaseisanalogoustoasmallopeneconomy with a stochastic endowment of tradable goods.

27 Price collapse after a long period of high prices I Experiment: we date the start of the high commodity price regime in the 2009:4 and its end on 2014:3. The low oil price regime goes from 2014:4 to 2016:1. We assume that agents have uninformative initial priors and experience the high-oil price regime during five years, followed by a year and a half of low prices. I Case 1: the sequence of revenues is {p t x t } T t=1 with {x t = d} T t=1.thiscaseisanalogoustoasmallopeneconomy with a stochastic endowment of tradable goods. I Case 2: commodity extraction is endogenous and responds to the random movements of commodity prices and we take the sequences {p t } T t=1 and { x t (p, s)} T t=1.

28 Price collapse after a long period of high prices in an endowment economy 0.08 Cond. Exp Oil Revenue 1.5 NT Price GDP 0.4 T Consumption Debt to GDP 0.3 Current Account to GDP RE BL SS

29 Price collapse after a long period of high prices in an extraction economy 1.5 Expected Oil Price 0.1 Extraction Reserves Exp. Oil Revenues 1.5 NT Price 3 Debt to GDP GDP 0.4 T Consumption Current Account to GDP RE BL SS

30 Final Remarks I We presented a framework to analyze the interaction between uncertainty about commodity price fundamentals and financial frictions.

31 Final Remarks I We presented a framework to analyze the interaction between uncertainty about commodity price fundamentals and financial frictions. I Model of natural resource extraction, incomplete financial markets and endogenous borrowing constraints capture macro dynamics (calibrated to Colombia and its oil sector)

32 Final Remarks I We presented a framework to analyze the interaction between uncertainty about commodity price fundamentals and financial frictions. I Model of natural resource extraction, incomplete financial markets and endogenous borrowing constraints capture macro dynamics (calibrated to Colombia and its oil sector) I We showed how discrepancies between initial expectations about prices and actual and posterior expected prices can be an important source of macro and financial instability because it affects non-trivially both households and firms.

33 Final Remarks I We presented a framework to analyze the interaction between uncertainty about commodity price fundamentals and financial frictions. I Model of natural resource extraction, incomplete financial markets and endogenous borrowing constraints capture macro dynamics (calibrated to Colombia and its oil sector) I We showed how discrepancies between initial expectations about prices and actual and posterior expected prices can be an important source of macro and financial instability because it affects non-trivially both households and firms. I Work in progress: I I I I Analyze the quantitative effects of alternative beliefs Shocks to R may be additional sources of uncertainty Oil ownership and operations may matter Financing extraction operations

34 Sources of information I Oil reserves, oil production (thousand barrels per day): US Energy Information Administration dataset (EIA). Annual data from 1980 to 2013 I Brent spot oil price (USD per barrel): US Energy Information Administration dataset (EIA). Annual data from 1980 to I Total public debt to GDP: World Development Indicators tables (WDI) and World Economic Outlook database (WEO). Annual data from 1979 to I Net Foreign Assets: Lane and Milesi-Ferreti (2007). Annual data from 1970 to I Default data: Borensztein and Panizza (2006). Annual data from 1979 to I GDP: World Economic Outlook database (WEO). Annual data from 1979 to 2010.

35 Oil price upswings and downswings Downswings Upswings Period Number of Months Period Number of Months NOV 75 - OCT NOV 78 - JAN FEB 81 - JUL AUG 86 - JUL AUG 87 - NOV DEC 88 - OCT NOV 90 - DEC JAN 94 - OCT NOV 96 - DEC JAN 99 - SEP OCT 00 - DEC JAN 02 - JUL AUG 08 - MAY TOTAL 219 TOTAL 196

36 Oil price swings and macro performance of net oil exporters 4 Real GDP Growth Rate (percentage points) 1 Current Account (percent of GDP) Real GDP Growth Rate (percentage points) Current Account (percent of GDP) Oil Production Growth Rate Upswing Downswing Oil Production Growth Rate Change in Net Foreign Assets (percent of GDP) Upswing Downswing Change in Net Foreign Assets (percent of GDP) 4 Upswing Downswing Upswing Downswing 4 Each bar is calculated with the same methodology of figure 4.4 of the World Economic Outlook of April of For default events the average number of default events is calculated.

37 New Title I Item 1 I I item a item b

38 New Title I Item 1 I I I Item 2 I I item a item b item a item b