Analysis of Interventions to Stabilize the Artemisinin Market

Transcription

Analysis of Interventions to Stabilize the Artemisinin Market
Analysis of Interventions to Stabilize the
Artemisinin Market
Burak Kazaz, Syracuse University
Scott Webster, Arizona State University
Prashant Yadav*, University of Michigan
8th Artemisinin Conference 2014
Guangzhou. China. Sep 23-24, 2014
* Presenting today
Background
• Cycles of bust and boom
• Changing ACT demand
patterns
• Semi-synthetic
artemisinin supply
• Multiple attempts to
stabilize the market in
the past
Source:
Prices upto 2012 are as reported at the Artemisinin Conference in
Hanoi in November 2011 by A2S2
Prices for 2012, 2013 and 2014 are 12 month average of monthly
median prices as estimated by WDI from India Import/Export
data
• Varying opinions and
perspectives on potential
ways to address the
challenges
Our approach
Research questions
• Impact of potential interventions on artemisinin (and ACT)
availability, price and volatility of price
• What types of interventions show promise, & what types do
not?
Approach
• Develop a mathematical model to capture essence of behavior
• Calibrate the model based on data gathered from field
• Use the model to understand the impact of different
interventions
Model
Key features to model
• Supply uncertainty
• Supply of semi-synthetic artemisinin
• Demand uncertainty
• Price volatility
• Farmer decision on what to grow
• Forward contracts
Enough richness to model a relatively complete range of supply
chain interventions, but not too much (“essence of behavior”)
Big picture view of the model
Level 2
Suppliers
Growing Season
Level 1
Manufacturers
Synthetic
supply
% of capacity
for alternative
Produce
artemisinin?
% of capacity
for artemisinin
ACT supply
Random
yield
ACT Market
Random
supply &
demand
Market clearing price
Expected
utility
from
alternative
Expected
utility from
artemisinin
Mean,
variance of
artemisinin
price
Forward contacts
(at expected
price)
Limitations
• Inventory carry-over effects are not modeled
• But, impact may be moderate, e.g., performance measures
relatively insensitive to increase in outside supply
• Assume full & symmetric information- Impact of
more accurate demand forecasts may be greater under
information asymmetry
• Do not capture the costs of interventions, only their
impact
• Identification of means and costs to affect different
parameters is left to those with specialized expertise
Model details- Supply
q = farm space growing Artemisia
2 = average yield per unit of
farm space
Z2 = supply random variable
E[Z2] = 1
2 = standard deviation of Z2
s = synthetic artemisinin supply
q2Z2 = (random) agricultural
artemisinin supply
q2Z2 + s = (random) total
artemisinin supply
Model details-Decision on what to grow
p  q   average price
 P2  q   price variance
ua = utility per unit of farm space
from producing artemisinin (by
“representative supplier”)
Ub = random utility of best alternative
b(u) = cdf of Ub
b = E[Ub]
 = supplier risk aversion parameter  0
=
2   p  q    P2  q  
b(ua) = % of farmers who prefer to
produce artemisinin
= P[Ub  ua]
Model details-space under forward contract
Ub0 = random utility of best
alternative for a unit of space
under contract
b0(u) = cdf of Ub0
ua0 = utility per unit of farm space under
contract (no price uncertainty)
= 2  p  q 
b0(u) = P[Ub0  u]
= P[Ub  u | Ub  ua0]
= b(u)/b(ua0)
i.e., space under contract is
representative of the population
Model details-farm space under forward contract
 = fraction of q under
q(1 – ) = space not under contract
contract
c = units of space that
could produce
artemisinin
= noncontract units producing
artemisinin
qb0(ua) = % under contract with Ub  ua
cb(ua) – qb0(ua) = units not under contract
with Ub  ua
= noncontract units producing
artemisinin
Equilibrium condition
q(1 – ) = cb(ua) – qb0(ua)
Demand Model M1
• p = price per unit of artemisinin
• E.g., ACT price = p + production cost
• M1: % of market willing to pay price p is independent of
the market size
• Motivated by a setting where the market is composed of
many individuals who buy ACT if willing/able to pay for it
Demand model M1
1 = expected market size
1Z1 = random market size
Z1 = demand random variable
d(p) = 1Z11(p) = random demand
E[Z1] = 1
market clearing price (supply = demand)
1 = standard deviation of Z1
q2Z2 + s = 1Z11(p)
= demand coefficient of variation
P(q) = random clearing price
1(p) = % willing to pay price p or
more
p0 = artemisinin support-price
(potential)
 1 
 q 2 Z 2  s   
 max  1  min 
,1  , p0 
 1Z1
  


Model
Demand model M2
• M2: Total volume sold at price p is independent of the
market size but depends on total budget
• Motivated by a setting where the market is composed of a
few large buyers willing to spend a fixed budget, b in a given
year
• Reality may be between M1 & M2, e.g., elements of both
Model
Demand model M2
b = total global spend (translated
into artemisinin equivalent)
d(p) = b/p = demand (isoelastic demand)
market clearing price
q2Z2 + s = b/p
P(q) = random clearing price


b
 max 
, p0 
 q2 Z 2  s

Measures of performance
• Fraction of need satisfied (fill rate) – public welfare

 q*  2 Z 2  s  
  E  min 
,1
 1Z1


• Coefficient of variation of Price (Price Volatility)
• Total expected supply – manufacturer welfare & public welfare
 1  q*  2  s
• Supplier surplus – supplier welfare


2  q 
E  2 p  q*   U b
*
   p  q  
 b 2
*





 
1

E  2  p  q*    2  q*   U b
    p  q*    2  q* 

b
2
p
 






 

Numerical analysis: base-case
Parameters & functions
Numerical analysis: base-case
Main Findings
Impact of Potential Interventions on:
•
•
•
•
•
Artemisinin Supply (MT of Supply)
Fill Rate (% filled)
Average Price
Price Volatility (CV of Price)
Extractor and Farmer Revenue
(surplus)
21
More forward contracts?
22
Impact of forward contract % () on supply
Impact of forward contract % () on price and volatility
More synthetic artemisinin?
25
Impact of synthetic production (s) on supply
Impact of semi-synthetic on supply (some risky special cases)
• Overall supply increasing in s, but there is risk of decreasing supply
•
Change from M2 base-case: suppliers are risk neutral & yield uncertainty increased by
33%
Impact of semi-synthetic on price volatility
Price support for artemisinin?
29
Impact of a price support
• Organizations commit to purchasing artemisinin at
stated support-price p0
• Product is held & sold in future period when price
is high
• Price support of $360 : 30% increase in supply, 60%
decrease in price volatility
Impact of support-price (p0) on supply
Impact of support-price (p0) on price volatility
Try to reduce demand
uncertainty?
33
Sensitivity to demand uncertainty (1)
Sensitivity of price to demand uncertainty (1)
Higher yielding seeds?
36
Impact of increase in average yield (2)
Impact of increase in average yield (2) on price and volatility
Increased budget for ACTs?
39
Impact of increase in budget (b) on supply
Impact of budget (b) increase on price and price volatility
Summary of Main Findings
• Limited impact from interventions that . . .
• increase the use of forward contracts
• improve forecast accuracy
• Greater impact from
• Use of a support-price (and reserve)
• Increase average yield
• Increase semi-synthetic supply with cautious transition
• Information sharing across supply chain actors is
critical
Artemisinin Supply Estimation
8th Artemisinin Conference 2014
Guangzhou. China. Sep 23-24, 2014
Credits
 This was supported by UNITAID (as part of the larger
API Market Intelligence Project)
 Some additional analysis was funded as part of a market
dynamics grant from the Bill and Melinda Gates
Foundation.
 Nora Hotte @ WDI was involved in conducting
interviews, analysis and helping develop insights
44
Background
 Between April-July 2015, the William Davidson
Institute (WDI) conducted an analysis of artemisinin
supply based on data analysis & triangulation,
stakeholder interviews and field visits
 This presentation summarizes the findings from that
work.
45
Predicted artemisinin production - 2014
Limitations
• Production will depend on spot market prices at harvest
as Chinese extractors will buy more wild leaves if
demand/price is reasonably high
• The cultivation of Artemisia annua is highly decentralized
and it is often difficult to accurately estimate agricultrual
output
• Extractors selling to each other (e.g. China and Vietnam
), could cause some production figures to be partially
double-counted
46
Predicted agricultural artemisinin production - 2014
Global artemisinin production in 2014 is estimated between 179-341
MT; 30-50% lower than in 2013
Since the Artemisia annua harvest
occurs between July-October in China, it
is too early to obtain actual production
numbers from extractors
400
341
350
Metric tons
300
260
250
200
Harvest would be between 30-50% lower
than 2013. Two main factors cited were:
• Low market prices disincentivized
farmers to cultivate and harvest
Artemisia annua
179
• Excessive rain may result in lower
yield per hectare
150
100
Some (small) extractors in Vietnam and
East Africa have temporarily stopped
producing artemisinin
50
0
Low
High
Conservative
47
Semi-synthetic artemisinin
Semi-synthetic production and sales in 2014 will depend on prevailing market
conditions
Scenario
SSA produced
(MT)
Market conditions
Low
25
 Market prices for natural artemisinin stay below US
$350/kg
 Market prices for SSA are ≥ US$350
 Sanofi Aventis produces only for in-house use
High
60
 Market prices for SSA are less than natural
artemisinin
 Sanofi Aventis produces to current capacity
Conservative
35-40
 Market prices for SSA are close to natural artemisinin
 Sanofi Aventis is able to sell 10-15MT of SSA to other
manufacturers who are interested in using SSA to
hedge future supply risks
Currently, few finished product manufacturers (except Sanofi-Aventis) have regulatory
approvals and processes setup to use semi-synthetic artemisinin.
48
Remaining stock from 2013 harvest
Findings
Artemisinin extractors, suppliers and traders are holding >70MT of
artemisinin from the 2013 harvest
Limitations
• Only collected data from a sub-set of artemisinin extractors in
China, Vietnam, East Africa, Madagascar who are supposed to be
the largest producers
• No data from artemisinin traders; however, is unlikely that they are
holding large quantities
49
Flow of stock from 2013 harvest
Stockpiling artemisinin to take advantage of low spot market prices?
India import data indicates only a 10% increase in artemisinin imports as compared to
the same time period in 2013.
Artemisinin Imported (Jan-Jul) in MT
25
2013
97.42
2014
107.14
Metric tons
20
15
2012
2013
10
2014
5
0
Jan
Feb
Mar
Apr
May
Jun
Jul
50
Current artemisinin market prices
Current reported spot market prices are around $250/kg, which is less than the break-even
point for artemisinin extractors. Low prices could impact the quantity of Artemisia annua
harvested in 2014 as well as future plantings.
Indian artemisinin imports (Apr-12 to Jul-14)
25
Artemisinin (MT)
700
Price (USD) / Kg
600
15
Price (USD)
500
400
300
10
200
5
100
Source: India import/export data purchased from Infodrive
Jul-14
Jun-14
May-14
Apr-14
Mar-14
Feb-14
Jan-14
Dec-13
Nov-13
Oct-13
Sep-13
Aug-13
Jul-13
Jun-13
May-13
Apr-13
Mar-13
Feb-13
Jan-13
Dec-12
Nov-12
Oct-12
Sep-12
Aug-12
Jul-12
Jun-12
0
May-12
-
Apr-12
Artemisinin imports (MT)
20
51
Conclusions
1 Based on our research, we expect that there will be
sufficient supply of artemisinin (from previous year’s
stocks, agricultural supply and semi-synthetic
artemisinin) to cover demand for QAACTs in 2015
2
3
If market prices remain low in late-2014/early 2015, then
Artemisia cultivation is expected to decline further and
some extractors may exit
Without interventions to maintain agricultural supply or
additional semi-synthetic artemisinin, a significant
decline in cultivation in 2015 could lead to an artemisinin
shortage and/or high market prices in 2016/2017
52