The Canadian and the United States

1

• The Canadian and the United States
• Softwood Lumber Market Integration
• A Comparative Perspective for the Periods of
• Softwood Lumber Agreement and Import Tariff
• Chander Shahi
• Fiona Yang
• Ph.D. Students
• University of Toronto
• Supervisor
• Dr. Shashi Kant

2
Outline

• Softwood lumber agreement
• Law of one price
• Earlier studies
• Data description
• Johansens multivariate cointegration tests
• Results for aggregate softwood lumber data
• Results for homogeneous products data
• Welfare impacts of SLA
• Conclusions

3
Softwood Lumber Agreement

• The period 1982 to 2004 saw a number of disputes
  in softwood lumber trade between the US and
  Canada
• Memorandum of Understanding in 1986
• imposed a 15 tax on certain softwood lumber
  bound for the US market
• Softwood Lumber Agreement, SLA (1996 and 2001)
• Restricted exports from four Canadian provinces
  to the United States
• voluntary export restraint agreement
• boosted US production
• In 2001 US imposed the CVD and AD duties (approx.
  27)

4
Law of One Price

• A single price for softwood lumber across Canada
  and US
• If the markets involved are integrated, these
  have to be modeled simultaneously
• Changes in prices of lumber in one region will
  lead to changes in other regions i.e. the markets
  are not independent
• Efficient arbitrage and continuous trade
  activities ensure that prices of softwood lumber
  sold by different producers in a market tend to
  uniformity

5
Earlier studies

• For North American softwood lumber markets
• Uri and Boyd (1990) tested the LOP for US lumber
  markets
• Jung and Doroodian (1994) tested the spatial
  equilibrium among four US regional lumber markets
  
• Catimel et. al. (1997) studied the trends in
  structure of the North American market for
  softwood lumber
• David Nanang (2000) tested the LOP for Canadian
  Softwood Lumber markets
• Runsheng Yin (2002) tested the LOP for US
  softwood lumber markets using disaggregated price
  series
• Runsheng Yin (2002) tested market integration
  among southern pine regions

6
Earlier studies

• For forestry products
• Boungiorno (1992) examined the LOP in US pulp and
  paper markets
• Hänninen (1998) tested the spatial equilibrium
  for sawn wood imported to UK from Finland and
  Sweden
• Prestemon and Holmes (2000) tested the spatial
  equilibrium relationship between US timber
  markets
• For export of forestry products
• Sarker (1996) studied Canadian softwood lumber
  export to the US
• Alavalapatti (1997) studied the exports of
  Canadian wood pulp to US

7
Markets selected
BC British Columbia PR Prairies ON Ontario QB
Quebec AT Atlantic Provinces
WC West Coast CA West coast of California WI West
Inland SO South NE North East
8
Data description

• Five Canadian softwood lumber markets (CANSIM)
• Atlantic Canada (Prince Edward Island, New
  Brunswick, Newfoundland and Nova Scotia)
• Ontario
• Quebec
• Prairies (Manitoba, Saskatchewan and Alberta)
• British Columbia
• Five US softwood lumber regional markets
• (Random Lengths yearbook 2004)
• West Coast region (Oregon and Washington, west of
  cascades)
• West Inland region (East of cascades in
  Washington and Oregon, Idaho, Montana and other
  western states)
• Southern region (The south from Virginia to
  Texas)
• West California Coastal region
• North Eastern region

9
Data description

• Homogenous products
• Douglas Fir (Madison Canadian Lumber reporter,
  2004)
• BC Coast (2x4 Std. Btr.) FOB mill prices
• BC Coast (2x4 Std. Btr.) delivered prices in US
  Atlantic Northeast
• BC Coast (2x10 2 Btr.) FOB Mill prices
• BC Coast (2x10 2 Btr.) delivered prices in US
  Atlantic Northeast
• Douglas Fir (Random Lengths yearbook, 2004)
• Portland (2x4 Std. Btr.) FOB mill prices
• Portland (2x4 Std. Btr.) delivered prices in US
  Atlantic Northeast
• Portland (2x10 2 Btr.) FOB Mill prices
• Portland (2x10 2 Btr.) delivered prices in US
  Atlantic Northeast
• Spruce Pine and Fir (Random Lengths yearbook,
  2004)
• Southern SPF
• Western SPF
• Eastern SPF(Delivered Great Lakes)
• Eastern SPF(Delivered Toronto)

10
Testing Procedures

• Cointegration
• A vector of variables is said to be cointegrated
  if each variable in the vector has a unit root in
  its univariate representation, but some linear
  combination of these variables is stationary
  (Engle Granger, 1987)
• Dickey Fuller test on cointegration regression
  residual
• The cointegration regression Durbin Watson test
  
• The Park J1 superfluous canonical cointegration
  regression
• The Hansen fully modified regression estimator Lc
  test
• The dynamic ordinary least square procedure
  (Stock Watson, 1988)
• The maximum likelihood cointegration approach
  (Johansen, 1988, 1991)

11
Johansens cointegration procedure

• Johansen maximum likelihood estimation
• Rank test defines the no. of cointegration
  vectors, r
• The rank is determined by estimating a
  p-dimensional VAR(k) model system
• Xt A1 Xt-1 -------------------------- Ak
  Xt-k u Dt et
• The re-parameterized equation in error correction
  form
• ?Xt ?1 ? Xt -1 --------------------- ?k-1 ?
  Xt k1 ? Xt-k u Dt et
• ?i -(I A1 - -------------------- - Ai),
  i 1,2, --------------, k-1
• ? -(I A1 - -------------------- - Ai)
  aß
• Likelihood ratio based trace tests
• HO r r0 HA r0 lt r lt p
• If r p-1, then the LOP holds for all prices
  simultaneously
• If r lt p-1, then the LOP is rejected for all
  prices simultaneously, in which case the second
  testable hypothesis is that the LOP holds region
  wise

12
Augmented Dickey Fuller Tests

• Verify the integrating properties (unit root
  tests)
• A variable Xt is said to have a unit root in
  its autoregressive process if it has the
  following univariate representation
• (1 - L ) Xt ?1(1 - L ) Xt-1 --------------
  ?1(1 - L ) Xt-pet
• A unit root examines if the variables are
  integrated of first order I(1)
• ADF test can be computed by running the following
  regression
• (1 - L ) Xt a ßXt-1 Spi1Fi(1 - L ) Xt-1

13
Results of ADF Tests

• SLA Period (1996-2001)

14
Results of ADF Tests

• Post SLA Period (2001-2004)

15
Appropriate lag length

• Akaike information criteria (AIC)
• Sims modified likelihood ratio test
• The test statistics is
• LR (T K) ln Dr - ln Du ?2(d)
• T - total sample size
• K - No. of variables in each unrestricted
  equation
• d - No. of restrictions
• Dr Du - The restricted and un-restricted
  covariance matrices
• We select a lag length of 3 for further analysis

16
Misspecifications tests of residuals

• Auto-correlation LM test for VAR models
• Heteroskedasticity ARCH test
• Normality Jarque-Bera test
• Standard errors Newey West test

17
Results of Johansens tests

• SLA period (1996-2001)

18
Results of Johansens tests

• Post SLA period (2001-2004)

19
Cointegration vector Weighting matrix

• SLA period (1996-2001)

20
Region wise comparison tests for LOP

• By imposing restrictions on the eigen values
  using a likelihood ratio test
• SLA period (1996-2001)
• Post SLA period (2001-2004)

21
Region wise comparison tests for LOP

• SLA period (1996-2001)
• Post SLA period (2001-2004)

22
Testing cointegration of similar products

• Douglas Fir
• BC Coast (2x4 Std. Btr.) FOB mill prices
• BC Coast (2x4 Std. Btr.) delivered prices in US
  Atlantic Northeast
• BC Coast (2x10 2 Btr.) FOB Mill prices
• BC Coast (2x10 2 Btr.) delivered prices in US
  Atlantic Northeast
• Portland (2x4 Std. Btr.) FOB mill prices
• Portland (2x4 Std. Btr.) delivered prices in US
  Atlantic Northeast
• Portland (2x10 2 Btr.) FOB Mill prices
• Portland (2x10 2 Btr.) delivered prices in US
  Atlantic Northeast
•  
• Spruce-Pine-Fir
• South (2x4 2 Btr.)
• Western (2x4 Std. Btr.)
• Eastern (2x4 1 2)
• Toronto (2x4 1 2)
•  

23
Results for similar products
24
Price Volatility studies

• Softwood lumber prices were more volatile in the
  1990s than in the 1980s (Zhang et al., 2001)
• Uncertainty and supply constraints due to the
  SLA have been assessed as the major reasons of
  price volatility
• Uncertainty about trade disputes continues
  between the two countries, so lumber prices
  continue to be volatile

25
Welfare impacts (Schwindt, 2004)
26
Conclusions

• The LOP doesnt hold for the Canadian and US
  softwood lumber markets
• Region wise LOP holds for the Canadian and US
  softwood lumber markets
• Market integration (No. of cointegrating vectors)
  increases during the post SLA period
• We need to ensure that there is a continuous
  trade between the markets under consideration
  before applying the Johansens cointegration
  tests
• For policy implications we need to study the
  efficiency of arbitrage between any two markets
  which take into account the transportation costs
• Since welfare impacts are different for different
  players in the trade theory, therefore, many of
  the policy actors would prefer a negotiated
  bilateral agreement