Docstoc

C fi Crude or Refined

Document Sample
C fi Crude or Refined Powered By Docstoc
					Crude R fi d?
C d or Refined?
Identifying Oil Price Dynamics through the Crack Spread and
Implications for Monetary Policy
I li ti      f M t        P li

                                               ,         ,
Seminar Presentation to Goldman Sachs Economics, New York, December 2010


Daniel P. Ahn June 1,
Insert Presentation Title 2011                                             June 1, 2011   1
Disentangling D  d dS      l Sh k
Di t li Oil Demand and Supply Shocks
»     Economists have begun appreciating that not all oil shocks are alike,
        d        i          it i il i       i                     lit f     th
      and assuming exogeneity in oil prices ignores reverse causality from the
      global economy through demand to prices.
»     Distinguishing between demand and supply-driven movements in oil
      prices has important normative as well as descriptive implications.
       » For example, a central bank may chose to raise interest rates in response to an
         endogenous demand-driven increase in oil prices but keep rates flat when faced with
         a contractionary cost-push supply shock.
»     Recent literature such as Kilian (2008) and Lippi and Nobili (2008)
      have relied upon a combination of price and q
                   p                      p               y
                                                   quantity measures to
      disentangle oil demand from supply shocks.
»     However, we argue that quantity measures may suffer from severe
                                               forward-looking
      measurement error and may not capture forward looking expectations
      of future shocks. Instead, we propose an alternative identification
      approach based purely upon price signals.


Information from Imperfect Pass-through in Oil Markets                               June 1, 2011   2
    Imperfect Pass-through in the Energy Complex
    I    f tP      th    h i th E        C   l
   »       End-use demand is for refined products, not raw crude oil. A demand
            h k has first-order impact on product price with a second-order
           shock h a fi t d i           t       d t i        ith        d d
           pass-through to crude prices. Similar logic applies for supply shocks.
   »       We extract the information from this asymmetric effect from imperfect
           pass-through to drive our disentanglement.
                          Market for Crude Oil                             Market for Refined Products




                                                                      hrough
 Supply                                                                                                             Demand
 Shock               First-Order Price Impact
                     First Order                                               Second-Order Price Impact
                                                                               Second Order


                                                      Imperfect Pass-th
                                                                                                                   Response


                                                              t
 Supply                                                                                                              Demand
Response                  Second-Order Price Impact                               First-Order Price Impact            Shock
                                                      I




    Insert Presentation Title                                                                                June 1, 2011   3
Agnosticism in Identification
A    ti i i Id tifi ti
»     We simply need that pass-through be imperfect for demand and supply
       h k to h                ti    d th f     di ti i h bl i       t
      shocks t have asymmetric and therefore distinguishable impacts upon
      the joint behavior of crude oil and product prices.
»     Borenstein, Cameron, and Gilbert (1997), Borenstein and Shepard
      (2002), and Davis and Hamilton (2004) argue that menu costs,
      drawdown of inventory buffers, seller market power, and incomplete
                                    pass-through
      information cause imperfect pass through of price shocks between
      crude and gasoline markets with a lag of 1-2 months.
»     No other strong assumptions, such as structural parameterizations, agent
      behavior, Ch l k t l               l d i                   Thi
      b h i or Cholesky-style causal ordering, are necessary. This
      identification approach is “agnostic.”
                                         Market Price Reaction
         Structural Shocks crude oil prices product prices crack spreads
         demand shock      ↑                ↑↑              ↑
         supply shock      ↑↑               ↑               ↓

Information from Imperfect Pass-through in Oil Markets                     June 1, 2011   4
Sign R t i ti A       h
Si Restriction Approach
»      We use two complementary methodologies to identify the
       disentanglement: sign restrictions and h t
       di t l           t i       t i ti             k d ti it id tifi ti
                                            d heteroskedasticity identification
»      The more intuitive approach uses sign restrictions in structural VAR
       identification as in Uhlig (2005), which Monte Carlo simulates multiple
       potential impulse response solutions of an under-identified VAR and
       averages those draws which ex post satisfy the imposed sign restrictions
»      Estimate following reduced-form Vector Auto-Regression




»                    keeping only those draws such that
                       » A>a>0 and B>b>0 or
                       » a>0, A-a>0, B>0, -(B-b)<0, and |B|>|-(B-b)|


Insert Presentation Title                                               June 1, 2011   5
Heteroskedasticity Identification
H t    k d ti it Id tifi ti
»      While intuitive, the sign restriction approach has some drawbacks.
       P h           t i      i th t th identification i    t i
       Perhaps most serious is that the id tifi ti is not unique and itsd it
       “impulse responses” are simply the average of multiple simulated
       responses, which may be dynamically inconsistent.
»      Hence, we also attempt the heteroskedasticity identification approach as
       in Rigobon (2005), which allows for conditional heteroskedasticity in
           reduced form
       the reduced-form shocks but assumes the parameters remain constant.
»      Specifically, we assume the primitive shocks follow GARCH(1,1)
       processes with conditionally uncorrelated innovations. Then we use
            i      lik lih d t     ti t th             ffi i t th t b t th
       maximum likelihood to estimate the VAR coefficients that best fit the
       GARCH processes.
»      Encouragingly, the uniquely identified impulses from heteroskedasticity
       identification are nigh indistinguishable from the sign restriction
       approach, suggesting our identification is robust


Insert Presentation Title                                               June 1, 2011   6
Impulse Responses from Sign Restrictions
I   l R           f    Si R t i ti




Insert Presentation Title                  June 1, 2011   7
Impulse Responses from Heteroskedasticity
I   l R           f    H t    k d ti it




Insert Presentation Title               June 1, 2011   8
        Real Crude Oil Prices from 1971
log prices
                                                               log Real Crude Oil Price                                               2008 oil price
    5                                                                                                                                 peak
                                     1980 Iran‐
                                     Iraq War

  4.5                                                                                                                     2003 US invasion
                                                                                                                          of Iraq
                          1979 Iranian
                                                                   1990 Gulf War
                          Revolution

    4                                                                                          1998 Russian
                                                                                               sovereign default
                                                                                               sovereign default


  3.5                                                                                                                              2005 Hurricane
                                                                                                                                   Katrina

                                                                                                                                      Great Recession
                                              1985 Collapse of                                                      9/11/01
    3                                                                                     1997 Asian 
                                              OPEC unity                                                            attacks
             1973 OPEC                                                                    Financial Crisis
             embargo

  2.5
     1971     1974        1977     1980      1983       1986        1989       1992       1995        1998         2001        2004      2007          2010
         Historical D     iti    D    d Shocks
         Hi t i l Decomposition: Demand Sh k
log real prices

  1.5



    1



  0.5



    0



 -0.5



   -1
                                      Cumulative Demand Shocks          Real Crude Oil Prices

 -1.5
    1971          1974        1977   1980    1983      1986      1989      1992       1995      1998   2001   2004     2007          2010




         Insert Presentation Title                                                                               June 1, 2011   10
          Historical D     iti    S   l Sh k
          Hi t i l Decomposition: Supply Shocks
log real prices

 1.5



   1



 0.5



   0



-0.5



  -1
                                      Cumulative Supply Shocks     Real Crude Oil Prices


-1.5
   1971           1974       1977      1980     1983      1986   1989      1992      1995   1998   2001   2004     2007          2010




          Insert Presentation Title                                                                          June 1, 2011   11
Random W lk down the History of Oil
R d    Walk d    th Hi t      f
»      Our decomposition of oil price dynamics are largely consistent with
       hi t i l i t iti (though in contrast with quantity measures!).
       historical intuition (th    hi      t t ith        tit         !)
»      The 1973 OPEC embargo, the 1979 Iranian Revolution, the 1986
       collapse of OPEC, the 1990 Gulf War, the 2003 US invasion of Iraq are
       all correctly classified as supply shocks.
»      Yet demand strength leading up to 1973 and also in 1979, followed by a
       large secular decline which troughed in 1998 during the Asian financial
       crisis, then a sustained recovery to 2007.
»      Interestingly, the 2008 “spike” is identified to be more supply-driven
       than demand-driven. Do we take the model literally and assume rational
       investors were expecting future supply shortfalls which ex post proved
       false? Or were prices distorted from demand/supply fundamentals by
       speculative pressures? But then must explain why speculation
       asymmetrically impacted crude over product markets.


Insert Presentation Title                                              June 1, 2011   12
    Decomposition of Variance Explained
    D       iti    fV i       E l i d
                       Oil Demand (% ) Oil Supply (% )
                                                          »   As one might intuitively expect,
horizon (months)       crude oil prices                       d      d h k in        d il i
                                                              demand shocks i crude oil prices
               1                      26.6         73.4       are smoother and more persistent
               6                      42.3         57.7
              12                     46.2          53.4       than supply shocks, which are
              24                     48.2          51.8       volatile and feature large tails.
              36                     48.3          51.7
              60                     48.4          51.7   »   Oil supply shocks drive most
horizon                crack spreads                          variation in crude oil prices in the
               1                      63.8         36.2       short-term, but both are import in
               6                      10.9         89.1
              12                       9.4         90.6       the long-term.
              24                      46 9
                                      46.9         53 1
                                                   53.1   »   M
                                                              More surprisingly, supply shocks
                                                                          ii l          l h k
              36                     48.7          51.3
              60                     48.7          51.3       drive nearly all variation in crack
                                                              spreads around the one year, but
                                                              demand drives variation both at
                                                              shorter and longer horizons.



    Insert Presentation Title                                                                June 1, 2011   13
Revisiting  Pi       d th M
R i iti Oil Prices and the Macro-economy
»      In an influential paper, Bernanke, Gertler, and Watson (1997) argued
       that d                  t     li     th th      il i
       th t endogenous monetary policy rather than oil prices per se was
       responsible for the 1970s stagflation.
»      Since then, economists continue to debate the relevance of oil price
       shocks on the variables such as output and inflation (Barsky and Kilian
       (2002) and Hamilton and Herera (2004)).
»                                                   question
       Our decomposition allows us to revisit this question, testing to see
       whether indeed oil demand shocks and oil supply shocks, as estimated
       by our identification schemes, have contrasting impacts on
                      i     i bl
       macroeconomic variables.
»      Indeed, we find demand shocks have positive and significant impacts on
       industrial production and core inflation, and negative on unemployment
»      By contrast, supply shocks negatively impact industrial production and
       positively impact unemployment.


Insert Presentation Title                                              June 1, 2011   14
                                   Response of IND to DMDOIL_H                             Response of IND to SUPOIL_H
                 .005



_
                                                                            .004
                 .004

                 .003                                                       .000

                 .002
                                                                            -.004
                 .001

                 .000
                                                                            -.008
                -.001

                -.002                                                       -.012
                            5       10    15     20    25        30   35            5       10    15     20    25        30   35


                                Response of CPI_CORE to DMDOIL_H                        Response of CPI_CORE to SUPOIL_H
                .0003                                                      .0003

                .0002                                                      .0002

                .0001                                                      .0001

                .0000                                                      .0000

               -.0001                                                      -.0001

               -.0002                                                      -.0002

               -.0003                                                      -.0003
                            5       10    15     20    25        30   35            5       10    15     20    25        30   35


                                Response of FEDFUNDS to DMDOIL_H                        Response of FEDFUNDS to SUPOIL_H
                   .3                                                          .2

                                                                               .1
                   .2
                                                                                0
                                                                               .0
                   .1                                                         -.1

                   .0                                                         -.2

                                                                              -.3
                  -.1
                                                                                4
                                                                              -.4

                  -.2                                                         -.5
                            5       10    15     20    25        30   35            5       10    15     20    25        30   35


Insert Presentation Title                                                                                                          June 1, 2011   15
                                 Response of UNEMP to DMDOIL_H                        Response of UNEMP to SUPOIL_H
                 .02                                                       .04



_                .01
                                                                           .03

                                                                           .02

                 .00                                                       .01

                                                                           .00
                -.01
                                                                          -.01

                -.02                                                      -.02
                            5       10    15    20    25     30    35            5       10    15    20     25    30    35


                                Response of CPI_CORE to DMDOIL_H                     Response of CPI_CORE to SUPOIL_H
              .0003                                                     .0003

              .0002                                                     .0002

                                                                        .0001
              .0001
                                                                        .0000
              .0000
                                                                        -.0001
              -.0001
                                                                        -.0002

              -.0002                                                    -.0003

              -.0003                                                    -.0004
                            5       10    15    20    25     30    35            5       10    15    20     25    30    35


                                Response of FEDFUNDS to DMDOIL_H                     Response of FEDFUNDS to SUPOIL_H
                  .4                                                        .2

                  .3
                                                                            .0
                  .2

                  .1                                                       -.2

                  .0
                                                                           -.4
                 -.1

                 -.2                                                       -.6
                            5       10    15    20    25     30    35            5       10    15    20     25    30    35


Insert Presentation Title                                                                                                    June 1, 2011   16
Smarter P li
S               ki           Sh k ?
        Policymaking vs. Oil Shocks?
»   As we argued, monetary and fiscal policymakers can and should respond
    differently to demand/supply shocks (if identifiable).
»   For a central bank, textbook theory suggests a demand-driven and oil
    shocks merits contractionary monetary policy. However, against a cost cost-
    push shock like an exogenous oil supply shock, the central bank faces a
    difficult trade-off between inflationary spillovers and lower output.
»   Interestingly, we do find evidence that the Fed Funds rate responded
    I t    ti l        d fi d id         th t th F d F d t              d d
    positively to an oil demand shock, but is ambiguous or negative to an oil
    supply shock.
»   At face value, during n the 2008 oil price spike, one may argue that the
    Fed was justified in holding off an interest rate hike (unlike the ECB)
    despite the threat of inflationary spillovers from higher commodity prices.
Reference Slides




Insert Presentation Title   June 1, 2011   18
Heteroskedasticity Identification under GARCH
H      k d i i Id ifi i             d
»   To identify the components of A0^-1, we first normalized the diagonal
    elements to (b_11,b_22) = (1, -1). We assume a GARCH(1,1) process for
    the reduced-form shocks ε_t, extracting residuals σ_it, i = 1,2.
»                                  (b 12,b 21)
    Then we search for the optimal (b_12,b_21) which maximizes the
    likelihood function:
        Measures of World Real Demand
log prices

  1.5
  15


    1


  0.5


    0


 -0.5


   -1
                                                 log real freight rates
 -1.5


   -2
    1971     1974   1977   1980   1983   1986   1989      1992        1995   1998   2001   2004   2007   2010
                   Response of DMDOIL_H to FGHT                        Response of SUPOIL_H to FGHT


       _
.06                                                         .12

.04
                                                             08
                                                            .08

.02
                                                            .04
.00
                                                             00
                                                            .00
-.02

                                                            -.04
-.04

-.06                                                        -.08
             5         10          15   20   25   30   35          5     10    15    20     25      30          35



                   Response of FGHT to DMDOIL_H                        Response of FGHT to SUPOIL_H
.20                                                         .20


.15                                                         .15


.10                                                         .10


.05                                                         .05


.00                                                         .00


-.05                                                        -.05
             5         10          15   20   25   30   35          5     10    15    20     25      30          35




       Insert Presentation Title                                                            June 1, 2011   21
        Measures of World Oil Production
log production

 0.15


  0.1


 0.05


    0


-0.05


 -0.1


-0.15


 -0.2
    1971         1974   1977   1980   1983   1986     1989       1992       1995    1998   2001   2004   2007   2010
                                               log detrended world oil production
                    Response of DMDOIL_H to PROD                         Response of SUPOIL_H to PROD


        _
 .06                                                          .12

 .04
                                                               08
                                                              .08

 .02
                                                              .04
 .00
                                                               00
                                                              .00
 -.02

                                                              -.04
 -.04

 -.06                                                         -.08
              5         10          15   20   25   30   35           5     10    15    20    25       30          35



                    Response of PROD to DMDOIL_H                         Response of PROD to SUPOIL_H
.020                                                         .020

.015                                                         .015

 010
.010                                                          010
                                                             .010

.005                                                         .005

.000                                                         .000

-.005                                                        -.005

-.010                                                        -.010
              5         10          15   20   25   30   35           5     10    15    20    25       30          35




        Insert Presentation Title                                                             June 1, 2011   23
Errors in Measures of Oil Supply
                                                 yoy
Official proved reserve levels are prone to huge y-o-y revisions
bn bbls y-o-y                         Saudi Arabia     Iraq             Iran
                                      Kuwait           UAE              Venezuela
130
                  UAE 1986: +64bn
110

  90                                                        Venezuela 2008: +73bn

  70

  50
                                      Saudi Arabia 1988: +85bn
  30

  10

-10
        1981                   1986     1991         1996       2001      2006
________
Source: BP Statistical Review, EIA
       Measures of Refinery Capacity Shocks
log capacity

  3%

  2%

  1%

  0%

 -1%

 -2%
  2%

 -3%

 -4%

 -5%
                                                  log Refining Capacity Shocks
 -6%
                                                                                          2005 Hurricane
 -7%                                                                                      Katrina

 -8%
    1971       1974   1977   1980   1983   1986      1989       1992      1995   1998   2001    2004       2007   2010
                                       Response of DMDOIL_H to LREF

_                           .04



                            .02



                            .00



                            -.02



                            -.04
                                   5     10    15     20    25    30   35


                                                   SUPOIL_H
                                       Response of SUPOIL H to LREF
                            .15


                            .10


                            .05


                             00
                            .00


                            -.05


                            -.10
                                   5     10    15     20    25    30   35

Insert Presentation Title                                                   June 1, 2011   26
Immediate and Expected Future Demand
I   di t    dE     t dF t     D    d
»      Using measures of global fundamental demand (in this case, shipping rates), we
       can decompose oil demand shocks into an immediate component and an
       expected future component. We find that shipping rates capture some but not all
       variation in our demand history




Insert Presentation Title                                                     June 1, 2011   27
Immediate and Expected Future Supply
I   di t    dE     t dF t     S   l
»      Measures of world oil production is very poor in capturing variation in supply
                                                                  Revolution,
       shocks , and notably the large shocks such as 1979 Iranian Revolution collapse
       of OPEC, 1990 Gulf War, and recent 2008 oil price spike are unexplained




Insert Presentation Title                                                     June 1, 2011   28

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:2
posted:10/8/2011
language:English
pages:28