The Cycle 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 - - PDF document

1

Long Live the Three-Year US Milk Price Cycle?

Chuck Nicholson

Department of Supply Chain & Information Systems

The Cycle

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

2006 2007 2008 2009 2010 2011 2012 2013 Average frequency 2006 – 2013: every 2.7 years

2

The Cycle

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

2006 2007 2008 2009 2010 2011 2012 2013 Chuck Talks about Price Cycles Average frequency 2006 – 2013: every 2.7 years

All-Milk Price Cycles (2008)

9, 26 and 36 month cycles $6.00/cwt variation Amplitude of cycles increasing?

3

All-Milk Price Cycles (2010)

Seasonal and cyclical components Largest cyclical component 36 month period

Other Cycles (2010)

Variable Range of Level Effect Amplitude of Seasonal Effect Largest Amplitude Cycle Amplitude

• f Largest

Cycle All-Milk Price $3.00/cwt $1.00/cwt 36-month $9.00/cwt Daily Milk Production 235 mil lbs 40 mil lbs 34-month 10 mil lbs Milk-Feed Price 1.2 0.5 33-month 1.0 Cheese Price $0.30/lb $0.20/cwt 36-month $0.80/lb Whey Price* $0.50/lb $0.05/lb 34-month $0.09/lb Class III Price $4.00/cwt $1.60/cwt 37-month $8.00/cwt NDM Price $0.65 $0.10/lb 34-month $0.70/cwt Butter Price $0.75 $0.20/lb 36-month $0.70/lb Class IV $4.00/cwt $2.00/cwt 34-month $8.50/cwt

*Since 2000

Indicates cyclical component large relative to range and(or) seasonal Rough convergence of periods of largest cycles

4

What’s New?

Context is now different:

• Rabobank says “the cycle is dead” due to

trade linkages

– Dysrhythmia (October 2012)

• Did the 2009 shock “re-set” or eliminate

the cycles?

• World of higher feed prices and new

business models

This I Believe…About Milk Price Cycles

Which represents your view?

• A. They never existed / evidence not

sufficient

• B. They existed but now they are dead
• C. They existed but I’m not sure what the

future holds

• D. They existed and probably will in the

future

5

Today’s Questions

• What does updated statistical analysis

indicate about the cycle?

• What causes cycles?

– Caused, if dead

• If the (a) cycle is still with us, what are the

implications?

Methods

• State-space (decomposition) statistical

analysis of All-milk price

– Level (average) – Slope (trend) – Seasonal (within year cycle) – Cycle

• Controls for effect of feed cost

– Also tested for effects of trade value/volume

6

Methods

• Quarterly data 1996(2) to 2012(1)

– 17 years, post-URA – All-milk price and NASS 16% protein ration

• Compare forecast to actual data for

2012(2) to 2013(1)

– If consistent, suggestive of continued cycle to date

Key Findings

• Slope, seasonal and cyclical factors

important

• Feed contributes $0.77/cwt for every $1

change in ration value

• Trade variables have limited impact

7

Model Testing

• Model passes all the usual statistical tests
• n errors (residuals)

– Normal, not serially correlated, homoskedastic

Predicted versus Actual 2007-2012

8

All-Milk and Feed Impact

5 10 15 20 25

1996(2) 1998(2) 2000(2) 2002(2) 2004(2) 2006(2) 2008(2) 2010(2) 2012(2) $/cwt

All Milk Feed Effect Feed has an impact on prices, but not on existence or timing of cycles

Seasonal and Cyclical Effects

• 6
• 4
• 2

2 4 6

1996(2) 1998(2) 2000(2) 2002(2) 2004(2) 2006(2) 2008(2) 2010(2) 2012(2) $/cwt

Seasonal Cycle Cyclical pattern exists through 2012, but peak of recent cycle lower

9

Cycle Findings

• Cycle length = 3.2 years

– 38 months

• Recent amplitude estimate = $1.50/cwt

– Less than previous cycle

Ex Post Forecast

Observations for 2012(2) to 2013(1) broadly consistent with model forecast

10

Price & Production Cycles Are Very Common

• Agricultural commodities
• Other commodities
• Housing
• GDP

Hog Production Cycle

• 1.5
• 1.0
• 0.5

0.0 0.5 1.0 1.5 2.0 2.5

1988 1992 1996 2000 2004 2008 2012

% Difference from Trend

11

Wheat Price Cycle

• 2.0
• 1.5
• 1.0
• 0.5

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

1970 1974 1978 1982 1986 1990 1994 1998 2002 2006 2010

% Difference from Trend

Housing Price Cycle

• 25.0
• 20.0
• 15.0
• 10.0
• 5.0

0.0 5.0 10.0 15.0 20.0 25.0

1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

% Difference from Exponential Growth

12

GDP Cycle

• 20.0
• 15.0
• 10.0
• 5.0

0.0 5.0 10.0 15.0

1947q1 1955q1 1963q1 1971q1 1979q1 1987q1 1995q1 2003q1 2011q1

% Difference from Exponential Growth

What’s in Common?

• Oscillations
• What causes oscillations?
• In systems speak:
• “Negative feedback loops with a delay”

– System structure causes behavior – This is the ONLY “system structure” that creates oscillatory behavior

13

Common Modes of Behavior

Exponential Growth

Time

Goal Seeking

Time

Oscillation

Time

S-shaped Growth

Time

Overshoot and Collapse

Time

Growth with Overshoot

Time

Oscillation is a very common behavior—it is caused by an underlying structure

Structure for Oscillations

State%of%the System Desired%State Discrepancy Correc4ve Ac4on

7 + + +

Measurement, Reporting and Perception Delays Administrative and Decision-Making Delays Action and Effect Delays

14

Structure for Milk Price Oscillations

(One structure among many possible ones)

Farm Profitability Desired1Farm Profitability Discrepancy Adjustment1of Cow1Numbers and1Feeding

= + + +

Measurement, Reporting and Perception Delays Administrative and Decision-Making Delays Action and Effect Delays

In Supply Chain Speak…

• Instability is common in supply chains
• Often the result of individual businesses

responding rationally to incentives

• But with delays and without sufficient

coordination

15

Have You Played “The Beer Game”?

• Game played to

simulate supply chain

• rdering decisions
• Often used with

groups of top executives

• Usually results in

instability--oscillations

Inventory Amounts Vary a Lot!

Actual results from The Beer Game with Executives

16

Orders vary and often amplified as they move upstream in the supply chain

This Instability is Called the “Bullwhip Effect” Why Cycles?

• Interaction of physical delays in production

and capacity adjustment with boundedly rational decision making by individual producers and investors

• Persistence of cycles suggests that

learning and market forces that might stabilize cycles are weak

17

What Do Models Suggest About Future Cycles?

• Forecasting three years ahead with State-

space model

– Model ignoring feed prices (so don’t have to forecast these)

SSM Suggests Continued Dampened Cycles

18

What Do Models Suggest About Future Cycles?

• We also modified our structural dynamic

model

– Updated to 2011 base data – Incorporates these supply chain effects

• Suggests continued cyclical behavior

– Although dampened by assumed lower feed price values based on USDA forecast

Dynamic Model Projections

15 16 17 18 19 20 21 22

1-13 1-14 1-15 1-16 1-17 1-18 1-19 $/cwt

Somewhat larger amplitude (SSM uses a dampening factor)

19

The Bottom Line

• There is evidence that cycles are not dead

– But they may be dampened in the future?

• Cycles arise from rational decisions by

supply chain actors

– Especially on the supply side?

• If cycles exist, forecasts and policy

analyses should account for them

The King is Dead…

• Long live the King!

20

Must These Cycles Exist in Dairy?

• 40
• 30
• 20
• 10

10 20 30 40 50 60

1991 1995 1999 2003 2007

Annual % Price Change

Price change behavior differs for different countries—one much more cyclical