WPS 13-10-2
Working Paper Series
Asymmetry in price transmission
in agricultural markets
Alain McLaren
October 2013
Asymmetry in price transmission in agricultural
markets.
Alain McLaren∗
Forthcoming in the Review of Development Economics
September 2013
Abstract
This paper explores the asymmetries in price transmission from international to local markets. We expect the presence of large intermediaries in agricultural markets to
lead to a stronger price transmission when international prices decline than when they
rise. The empirical evidence conrms the presence of asymmetric price transmission
consistent with the presence of large intermediaries with monopsony power.
JEL codes: Q13, Q17.
Keywords: Asymmetric price transmission, agricultural markets.
University of Geneva, Département d'Economie Politique, Unimail, Bd. du Pont d'Arve 40, 1211 Genève,
Switzerland. Telephone: +41-788431321. Email: [email protected].
I am grateful to Jean-Louis Arcand, Richard Baldwin, Marcelo Olarreaga, Frederic Robert-Nicoud,
Takamune Fujii and Henry Kinnucan for their helpful comments and discussions. I would also like to
thank all participants at the "ProDoc Trade PhD Workshop", the "University of Geneva Young Reseachers
Seminar" as well as the "Bari Third International Workshop on Economics of Global Interactions: New
Perspectives on Trade, Factor Mobility and Development" for their useful comments and suggestions.
∗
1
"While grocery store shelves appear to provide abundant choices, most of these
products are marketed by a small and decreasing number of rms. Gigantic multinational corporations are consolidating their control over our food system [...].
The trend raises concerns about how this power is exercised, as most of these
corporations are accountable to their shareholders, not to communities in which
they operate." Phil Howard (Howard (2006))
1
Introduction
Many poor countries have a large proportion of their active population working in agriculture.
In many cases, the amount of revenue they receive will be crucial for their survival. Using
data from the World Development Indicators (2010) one sees that in the year 2000, the world's
25 percent poorest countries had at least a quarter of their population working in agriculture,
and half of these 25% poorest had over 70% of their population active in agriculture. An
illustration of this is given in Figure 1. Thus poorer countries will be more exposed to large
falls in agricultural prices. Understanding the determinants of changes in agricultural prices
is therefore crucial for the poorest countries.
In this paper the focus is on the extent of asymmetry in the price transmission from international to local markets. Whether a farmer is selling locally or exporting, the price he will
receive for his production will be directly or indirectly aected by prices determined in world
markets. Indeed, Mundlak & Larson (1992) show that variations in local agricultural prices
are mainly explained by variations in world prices. But the transmission from international
to local markets may not necessarily be symmetric. Depending on market conditions falls
in international prices may be better transmitted to local markets than increases in international prices. The consequences of this asymmetric price transmission could be particularly
harmful in poor countries where farmers often live close to the poverty line. In fact, Mosley
& Suleiman (2007) suggest that a portion of the small farmers are below the poverty line
2
in some of the poorer countries. They also put forward that the one sector that has had a
strong ability to stimulate pro-poor growth processes, especially in East and South Asia, is
smallholder agriculture.
Why would one expect a better price transmission when agricultural prices fall? Agricultural markets are characterized by the presence of large international intermediaries, with
strong monopsony power over often small and numerous producers. Murphy (2006) shows
that in the United States two companies (Cargill and Archer Daniels Midland) export 40
percent of all U.S. grains. Rogers & Sexton (1994) show that in the United States more
than 60 percent of all food and tobacco markets can be considered as noncompetitive when
measured by their top four-rm concentration ratio (with a threshold at 50 percent). Figures
for other countries are similar. For example in Vink & Kirsten (2002), concentration ratios
for the four largest rms for South Africa are also large: 47 percent in slaughtering, dressing
and packaging livestock, 65 percent for vegetables and animal oils and fats, 43 percent for
our, 37 percent for animal feeds, 99 percent for sugar, golden syrup and castor sugar, 80
percent for coee, coee substitutes and tea.
In this paper it is shown that in the presence of strong monopsony power of agricultural
intermediaires with suciently convex cost functions one should expect an asymmetric price
transmission which is consistent with the use of this monopsony power by intermediaires.
Indeed as international price falls, local prices will fall proportionally more than when international prices increase. This prediction is conrmed when confronted to a sample of
161 agricultural products produced in 117 countries over a period of 35 years. Moreover,
the asymmetry seems to be driven by the results for markets where large international intermediaries are present or when exports represent a large share of total production which
increases the monopsony power of international intermediaries.
Questions of asymmetric price transmission have been widely studied for oil markets,
known as the literature on "Rockets and Feathers", where prices rise like rockets but fall like
feathers. Many researchers have been analyzing the evolution of oil output prices. Tappata
3
(2009) analyzed the theoretical aspect of potential asymmetric responses of oil retail prices.
Many empirical studies come to the conclusion that asymmetry does exist. This is the case
of studies such as Galeotti et al. (2003) or the main study on the topic done by Borenstein
et al. (1997). The latter has several explanations for asymmetry. The one than seems to
be closest to what is found for agriculture is one of costly search of consumers. The idea is
that the consumer will believe that a change in price at the retail station during a period of
volatile crude oil prices will really be due to a change in this price, whereas in less volatile
periods he will believe that the station's margin is changing. He will put more eort into
searching for a lower price elsewhere if he believes that it is that specic station increasing its
price. This is because the consumer's expected gain of search is higher in this case. This, in
turn, leads to a higher market power of retailers who can dampen the rate of passthrough of
upstream price decreases. This paper shows similar eects for agriculture, however aecting
the producer side rather than the consumer side.
The results are interesting from a policy perspective and give room for active competition
policies when facing large international intermediaries. Levinsohn (1994) suggests that many
countries are more lax in their competition policy when dealing with export markets, perhaps because anticompetitive practices in an export market will not be harmful for domestic
consumers. However, the asymmetric price transmission identied in this paper would lead
to increased losses for often poor farmers. More generally, as argued by Murphy (2006),
market power in international markets is not factored into the models and assumptions that
inform the trade and agriculture debate, which can mislead policy makers in terms of the
distribution of the gains from trade liberalization in these markets. This result can also oer
some interesting perspectives in terms of aid-eectiveness. Some studies such as Mosley &
Suleiman (2007) have made clear some important aspects of aid, such as the importance of
food crops due to their high poverty leverage. The result of this paper concerning intermediaries may complement such studies, proposing some strategies that may be adopted within
the agricultural sector. An asymmetric price transmission could also alter the usual way of
4
approaching the analysis of welfare linked to price changes. De Hoyos & Medvedev (2011)
study the changes in welfare created by the price transmission from international prices to
domestic prices. They put forward the importance of the degree of price transmission in
explaining short to medium-term poverty eects. The presence of asymmetric transmission can give new insights to such theories and a new way of thinking about the eects of
international price changes for welfare.
The remainder of the paper is organized as follows: section 2 presents a simple model to
illustrate how asymmetric price transmission can arise in the presence of large intermediaries
with suciently convex cost functions. Section 3 presents the data used to empirically explore
this question and section 4 presents the empirical strategy. Results are presented in section
5 and section 6 concludes.
2
Theoretical background
Agricultural markets are characterized by a large dispersion of farmers, as noted by Sexton
(1990), which are numerous and therefore act as price takers. He also emphasizes the bulkiness and/or perishability of raw products that will have an inuence on market structure.
This could lead to what he calls spatial oligopsony power of processors or wholesalers. As
quoted in OECD (2008) it is the diculty for sellers to nd other buyers which determines
the extent of a buyer's monopsony power. Murphy (2006) says that most farmers lack the
storage and capital needed to get their goods to distant markets, so they are left selling
locally, to middle-men who now have more suppliers to choose from.
Market power is modeled based on Sexton (1990). Wholesalers are considered homogeneous and act as price takers in their selling markets, notably due to the size of international
markets. The xed cost associated with exporting is too high for the individual farmer to
face, such that he has to pass by a wholesaler. This is consistent with the literature, notably
with Gopinath et al. (2007) who say that agriculture is unique as farmers often do not export
5
directly since it is marketing rms that make the export decision.
The international price of a good is denote by P ∗ and is considered as being exogenous.
The wholesalers therefore face an innitely elastic demand and it is assumed that they benet
from economies of scale, such that there will be much fewer of them than farmers. Dening
the price received by the farmer as pf i , such that one can write the supply of one farmer as
in equation (1).
qis = qi (pf i )
with
qi (pf i )0 > 0 and qi (pf i )00 ≥ 0
(1)
By summing over all farmers one gets the aggregate supply of farmers to a processor,
presented in equation (2).
Qs = Qs (pf )
with Qs (pf )0 > 0 and Qs (pf )00 ≥ 0
(2)
For later use one can dene the inverse supply function as in equation (3).
(Qs )−1 (Q) = pf
(3)
This inverse supply will be called w(Q) which leads to the expression given by equation
(4).
w(Q) = pf
with w(Q)0 > 0 and w(Q)00 ≥ 0
(4)
The wholesaler must either package the products or incur some extra costs to export the
product. These costs will be increasing in the quantity of raw product Q and will be given
by the function m(Q). For higher quantities, the wholesaler will have to pay even higher
costs to get products from farmers that are further away.
Besides these costs the wholesaler must incur the cost of buying the product, w(Q)Q as
6
well as a xed cost to export. Total cost is given by equation (5).
c(Q) = w(Q)Q + m(Q) + f
(5)
The xed cost to export mentioned above is denoted by f and is added to the model such
that there are economies of scale. This will spread out the dierent wholesalers geographically, instead of having one at each farm site.
The next step is to depart from the analysis of Sexton (1990) to see what will happen if
there is an exogenous shock to P ∗ , as illustrated in Figure 2.
Due to the fact that an increased quantity must at least in part be supplied by farmers
that are further away, which comes on top of the usual increasing supply in the presence
of monopsony power, the marginal cost will be increasing and convex in Q and is given by
equation (6).
mc(Q) =
dw(Q)
∂m(Q)
∂c(Q)
=
Q + w(Q) +
∂Q
dQ
∂Q
(6)
The supply curve of the wholesaler on the international market is his marginal cost. On
Figure 2 one can see that a change in international price leads to a lower increase in quantity
than the decrease in quantity associated with a decrease in price of the same magnitude. This
larger decrease in quantity will induce a larger change in pf for a decrease in international
price if the farmers supply has a constant slope or if it doesn't, as long as marginal cost is
suciently convex, as will be presented below. This intuitive mechanism will be formally
developed below, with a model linking international price to producer price.
One can then formulate the wholesaler's prot function as given by equation (7).1
1I
owe special thanks to Henry Kinnucan for proposing a simpler and shorter route in nding the relationship between farmer price and international price. The development below, from equations (7) to (15),
closely follows the development that he proposed.
7
max π = P ∗ Q − w(Q)Q − m(Q) − f
(7)
In this model w(Q) = pf is the inverse supply function for the farm-based input and
m(Q) is the cost function associated with preparing the farm product for home consumption
or export, and f is the xed cost mentioned above.
By maximizing prot with respect to Q one gets the First Order Condition (FOC) as
presented in equation (8).
P ∗ − w0 (Q)Q − w(Q) − m0 (Q) = 0
(8)
where the quantity that optimizes prot is denoted Q. What we are interested in here is
how optimal quantity will vary for a given change of P ∗ . Since one can't directly isolate Q
in equation (8), an implicit function is used. Denoting F (Q) as being the left hand side of
the FOC leads to the identity presented in equation (9).
F (Q, P ) = P ∗ − w0 (Q)Q − w(Q) − m0 (Q) ≡ 0
(9)
The implicit function is the one given in equation (10), which is the wholesaler's supply
function.
Q = Q(P ∗ )
(10)
The use of the Implicit Function Theorem is now useful to dierentiate optimal quantity
with respect to price, as in equation (11).
8
dF
dQ
1
dP ∗
=
−
=
−
dF
dP ∗
−w00 (Q)Q − w0 (Q) − w0 (Q) − m00 (Q)
dQ
=
(11)
1
> 0
w00 (Q)Q + 2w0 (Q) + m00 (Q)
The relationship between farm and wholesale prices that was given in equation (4) can
now be written, at the optimum, as in equation (12).
(12)
pf = w(Q) = w[Q(P ∗ )]
Using the chain rule, the change of farmer price for a change in the international price is
given in equation (13).
dw(Q(P ∗ )) dQ
w0 (Q)
dpf
=
=
> 0
dP ∗
dP ∗
dQ
w00 (Q)Q + 2w0 (Q) + m00 (Q)
(13)
Two important remarks can be said concerning the relationship between international
price and farmer price. The rst is that the relationship is increasing, and the second is that
0<
dpf
dP ∗
< 1.
2
It basically means that a one dollar increase in wholesale price causes farm
price to rise by less than a dollar. If the farm supply and m(Q), the cost of packaging as well
as other extra costs to export, are linear functions, then
dpf
dP ∗
= 12 , and the one dollar increase
in international price mentioned above would lead to a 50 cents increase in farm price.
Any asymmetry in the price transmission will take place if the price transmission relation
is non-linear in wholesale price. This can be evaluated by taking the second derivative with
respect to P ∗ of equation (12). This is given by equation (14).
2 I thank Henry Kinnucan for pointing out that in much of the literature, the term "imperfect price
transmission" is used in situations such as this, which is not entirely correct. For a detailed explanation of
this, see Kinnucan & Zhang (2013) in which it is shown that a farm-retail price transmission of 1 isn't a
prerequisite for competitive market clearing and that one must distinguish between the elasticity of price
transmission and the slope of price transmission.
9
d2 pf
dQ
= w00
∗
2
(dP )
dP ∗
!2
+ w0
d2 Q
(dP ∗ )2
(14)
Asymmetric price transmission requires non-linearity of either the farm supply or the
wholesale supply functions. If both functions are linear, then
d2 pf
(dP ∗ )2
= 0, and price transmis-
sion is symmetric.
Whether the sign of the right hand side of equation (14) is negative or positive is indeterminate. Although w0 (Q) and w00 (Q) are positive as dened higher up, and the squared
term
dQ
dP ∗
2
> 0 is also positive, the last term
d2 Q
dP ∗2
is negative (assuming convexity of the
marginal cost curve, if it were linear then this term would be zero).
The sign of equation (14) can be made determinate by placing restrictions on the shapes
of the farm and wholesale supply functions. This becomes clearer when rewriting it with w00
on the left hand side, as in equation (15).
2
2
w00 <
d Q
−w0 (dP
∗ )2
dQ
dP ∗
2
∗
d P
w0 (dQ)
2
= dQ
dP ∗
(15)
2
One such restriction is that the inverse wholesale supply function must be suciently
convex compared to the inverse farm supply function. If this is the case, then there will be
a negative sign in equation (15).
For the wholesalers inverse supply function to be convex, that is
supply function must be concave, meaning that
d2 Q
(dP ∗ )2
d2 P ∗
(dQ)2
> 0, its ordinary
> 0. The latter is a necessary condition
for equation (14) to have a negative sign, and thus produce the concave function for the price
transmission relation shown in Figure 3.
To summarize, the relationship between P ∗ and w is increasing and concave as long as
the marginal cost curve is suciently convex with respect to the inverse supply curve of
farmers. If this is the case, the relationship between the two will be such that transmission
will be larger if there is a downward international price movement from the equilibrium than
10
if there is an upward movement.
3
Data
In order to estimate the impact of international agricultural price variations on producer
prices, yearly data on export and producer prices from the Food and Agriculture Organization of the United Nations (FAO) is used. The producer price is the price received by
farmers as collected at the point of sale for primary crops, live animals and livestock primary
products. FAO's export data is produced according to the International Merchandise Trade
Statistics Methodology and mainly comes from national authorities and other international
organizations. The export values are reported as Free-on-Board (FOB) such that insurance
and transport costs are not included. It is an unbalanced panel of 161 items, 117 countries
and 35 years, ranging from 1966 to 2000. The list of items and countries are presented in
the Appendix.
Higher frequency data are usually used in the Feathers and Rockets literature as well as
some studies on agricultural commodities' price transmission. However, the latter are usually
case studies, covering a few items or a specic country. This paper is aimed at a more general
approach. Tomek & Myers (1993) say that farmers often make annual decisions. Crops are
an example, where the farmer decides at the beginning of the period the area where he is
going to plant, then can't change it. Other similar decisions are taken by the farmers which
can't then be changed, whether international prices change or not.
The data used for the meteorological IV's comes from Mitchell et al. (2003). For the
natural catastrophies what is used is
The OFDA/CRED International Disaster Database
(n.d.). The neighbouring countries' data was taken from the
World CIA Factbook (2011).
As the interest here is on price transmission, exporters who are considered large are
excluded from the sample. A benchmark of 1% of total exports in each item-country pair is
set as a small exporter. This is to stay in line with the idea of a "small country" being a
11
price taker. This is also a rst step towards dealing with endogeneity.
Data is also cleaned in a way such that extreme values are excluded. Even though intuition leads to expect a price of exports above the domestic price, due to processing and
handling costs, we however allow an export price as low as 80% of the producer price in
order to include dierent possible scenarios, such as exceptional price variations. A ratio of
export price over domestic price that is above 20 is also excluded, as producer prices reported
as being less than 5% of the export price suggests that what is captured is probably that
the product may be slightly dierent from what is sold locally.3 Beyond product quality
dierences and potential mistakes in the data, the exclusion of export prices that are considerably lower than producer prices can also be justied by the fact that it may reect some
"dumping" mechanisms.4
To study the link between oligopsony and asymmetry, data on one of the main companies
that trades agricultural products worldwide is included. This data was collected by looking
at Cargill's worldwide website. The company reports each country in which it is present and
have either a page or a specic website for each. The information given includes the year
from when it has been active in that country and the products that it covered in that country.
Matches were done so that the product names were comparable to the FAO database. The
list of items covered by Cargill are presented in the Appendix.
3 There is a relatively large amount of observations for which the ratio is below 0.8. These cases don't
seem to be coming from any countries or items in particular, but the frequency does drop somewhat after
1990. This seems to be coming from the fact that FAO producer prices come from two databases, one being
considered as historical price data and intitled "Price archive". The latter is probably of lower quality than
the more recent data. Regressions without any cleaning were run but low R2 and Hansen test values conrm
that some cleaning is necessary.
4 According to the World Trade Organisation, if a company exports a product at a price lower than the
price it normally charges on its own home market, it is said to be "dumping" the product. We can therefore
beleive that if companies are engaged in dumping, the causality and mechanism of price transmission isn't
as clear.
12
4
Empirical strategy
4.1
Asymmetry of price transmission
International prices are proxied by export prices and the explained variable is producer
prices. The log of export price will be regressed on the log of producer price such that one
can interpret the price transmission coecient as an elasticity. A dummy taking a value of
1 if the export price for a certain good in a specic country increased from the previous year
and 0 if it decreased is added to the regression. This will enable us to distinguish increasing
prices from decreasing prices.
The specication used is given by equation (16). λic , λct and λit are respectively itemcountry, country-year and item-year xed eects.
prod
exp
ln(Pi,c,t
) = α + γ(price upi,c,t ) + β ln(Pi,c,t
)
(16)
exp
+δ(price upi,c,t ) ln(Pi,c,t
) + λic + λct + λit + i,c,t
Items are noted by i, with i = 1, ..., N , countries by c, with c = 1, ..., C and years by t,
with t = 1, ...T .
To get rid of the item-country, country-year and item-year xed eects the transformation
given by equation (17) is used.
prod
prod
prod
prod
prod
ln(Pi,c,t
) = ln(Pi,c,t
) − ln(Pi,c,·
) − ln(Pi,·,t
) − ln(P·,c,t
)
g
prod
prod
prod
prod
+ ln(Pi,·,·
) + ln(P·,i,·
) + ln(P·,·,t
) − ln(P·,·,·
)
with
prod
ln(Pi,c,·
)≡
1X
Pcit
T t
13
(17)
prod
ln(Pi,·,·
)≡
prod
ln(P·,·,·
)≡
1 XX
Pcit
CT c t
1 XXX
Pcit
CN T c i t
The other means being dened in the same way.
This transformation is applied to all variables individually, which will eliminate the xed
eects and the constant such that what is left is presented in equation (18).
prod
gexp )
g up) + β ln(P
ln(Pi,c,t
) = γ (price
t
i,c,t
g
(18)
g ln(P exp ) + g
+δ (price up)
t
i,c,t
i,c,t
The β coecient is the elasticity of price transmission. The asymmetry is given by the δ
coecient. A value dierent from zero will reveal asymmetry.
Results for the Ordinary Least Squares (OLS) estimation are given in column 1 of table 1.
Standard errors are clustered by item-country in this regression as well as for the Two-Stage
Least Squares (2SLS) regression presented below.
4.2
Dealing with endogeneity
As mentioned above, the export price may be endogenous. This implies that the price
up dummy and the interaction may also be endogenous. For this reason the method of
estimation that will be used is that of 2SLS. For this one must nd a variable that will
aect export price without however being directly correlated to producer price. This can
be done by using climatological phenomena in other countries exporting the same good.
Three meteorological Instrumental Variables (IV's) will be used, which includes the amount
of rainfall, temperature and cloud cover. A variable of catastrophic events is also used as
14
an instrument. To avoid any correlation between the weather conditions in the country
considered and the weather in other countries, all neighbours are excluded when creating
the instrument.
The three meteorological IV's will be created in the same manner. As it is hard to know
what the eect of a certain change of each variable on prices will be and since the study
contains many dierent types of goods, it will be considered that being far away from the
mean is bad for the farmer. This is close to Shaw (1964) where it is said that a reasonable
way of seeing the relationship between meteorological factors and yield is by representing it
as a bell-shaped curve. In our case the eect of the three IV's will be seen in this way such
that what will be used will be the standard deviation of the monthly temperature, monthly
rainfall and cloud cover each year.
What is considered is the eect of these variables on all other countries that are exporting
the good, multiplied by their share of exports with respect to the rest of the world in that
good, and excluding neighboring countries. This exclusion is to avoid any direct correlation
between the neighbouring countries' weather conditions and the producer price considered.
This will give us exogenous shocks to all other countries' prices, therefore inuencing international price but not the own country's price directly.
An illustration of the construction of these variables is given in Figure 4 for bananas in
Zimbabwe in the year 2000. This illustrates that all neighbours and non exporters of the
good aren't included in the instrument. All others are counted, the weight in the instrument
depending on their export share. This is explicited in equation (19).
rain IVb =
X
rainf allict ∗ export shareict , c 6= n and c 6= b
(19)
c
with: b: own country
c: country
n: neighbours
Finally, a count variable of the number of climatological (including extreme temperature,
15
drought, wildre) or meteorological (storm) events in other countries is added as another
instrument. It is constructed in the same way as before but this time the number of disasters
is used directly rather than the standard deviation.
As the price up dummy may also be endogenous, it is instrumented by a dummy taking
the value 1 if the IV variable in question went up and taking the value 0 otherwise. This
dummy is also interacted with the IV variable itself in order to instrument the interaction
of price up with the export price.
5
Results
Results of the regression using the IV's are given in column 2 of table 1. The results
of the rst stage regressions are given in table 2. The interaction term clearly points to
some asymmetry. The sign of the asymmetry term is the same as in the OLS regression
given in column 1 and the p-value of the test of underidentication as well as the one for
overidentifying restrictions suggest that the instruments are valid. Empirical results suggest
that when FOB export prices rise by 1% farm prices rise by 0.60%,5 and when FOB export
prices fall by 1% farm prices fall by 0.98%. In other words, there is near perfect transmission
of declines in wholesale price, but imperfect transmission of rises in wholesale price. This is
consistent with a situation in which the wholesalers supply function is suciently convex.
5.1
Robustness checks
The results above show that there is an asymmetry in the long run price transmission of
international prices to local prices. However, it is important to check the immediate response
as well. The inclusion of the lagged value of producer price as an explanatory variable will
enable one to do the aforementioned interpretation.
The specication used is given by equation (20). λi and λc are respectively item and
5 From
table 1 we can work out the upward tranmission as being 97.83% − 37.47% = 60.36%.
16
country xed eects.
prod
prod
exp
ln(Pi,c,t
) = α + ρ ln(Pi,c,t−1
) + γ (price up)t + β ln(Pi,c,t
)
(20)
exp
+δ (price up)t ln(Pi,c,t
) + λi + λc + i,c,t
First-dierencing will eliminate the xed eects as well as the constant. However, one
prod
then has ln(Pi,c,t−1
) on the left hand side which will be correlated with the error term i,c,t−1 .
To deal with this problem the estimation method used is that of Arellano & Bond (1991) and
Arellano and Bover / Blundell and Bond (Arellano & Bover (1995); Blundell & Bond (1998))
who developed a Generalized Method-of-Moments (GMM) estimator that instruments the
dierenced variables that are not strictly exogenous with all their available lags in levels
or in rst dierences. It is implemented using the approach of Roodman (2006) based on
the Arellano and Bond (Arellano & Bond (1991)) and Arellano and Bover / Blundell and
Bond (Arellano & Bover (1995); Blundell & Bond (1998)) dynamic panel estimators. As
mentioned above, the explanatory variables used here cannot be considered strictly exogenous
and the change in the export price cannot be considered uncorrelated with item and country
unobservable xed eects, therefore the original equation in levels is not added to the system.
The export price, the price change as well as the interaction term are considered as being
predetermined such that the rst lag that is used as an instrument for these variables in t is
t − 2. The rst lag for producer prices that is used in t − 1 (the lag of the explained variable)
is t − 3. When running the regression on the whole data the N dimension is composed of
countries and items and is therefore much larger than T . Even though this is good in the
case of these types of estimators requiring small T and large N , only half of the periods in
the sample are used as instruments to limit the number of instruments used. Time dummies
are also included in the regression.
As the suciency of the variables' past values being used as instruments may be questionable we also use a specication with the above mentioned economic instruments, used in
17
the 2SLS specication. The results are shown in table 3.
These results support the ones found in the OLS and 2SLS specications, that is a
negative and signicant asymmetry term. Due to the lag of producer price on the right hand
side of equation (20) the interpretation of the coecients is slightly dierent. The short
run price transmission is given by β in equation (20) whereas the long run transmission,
comparable to the β in equation (18), is computed as follows:
is obtained by computing ρ −
β−δ
.
1−ρ
β
.
1−ρ
The long run asymmetry
This yields the results presented in table 4. Standard
errors are computed using a calculation technique based on the "delta method".6
One sees that the initial transmission, although relatively small, is asymmetric. This
then leads to a long run transmission that is, albeit slightly smaller than what is found with
the 2SLS approach, not dierent in terms of statistical signicance.
5.2
Oligopsony and asymmetry
The next step is to see whether measures of market power are linked to asymmetry. For this
two variables are used. The rst is the importance of exports in the market. By looking at
the quantity exported with respect to the quantity produced, we will be able to see whether
asymmetry is inuenced by a larger share of exports.7 One might expect the fact of exporting
more of local production to lead to more price transmission. However, this would not explain
more asymmetry. What is of interest here is whether the mechanism specic to exporting
plays a role in explaining asymmetry. A larger local presence of intermediaries will be a
prerequisite for exporting a larger share of local production. A dierence in asymmetry for
high versus low shares of exports will point to the impact of intermediaries in the asymmetric
pattern of price transmission. We will therefore use two groups of countries, those exporting
a large share of their production and those exporting a small share. Results are given in
table 5. One sees that in markets where exports consist of a large share with respect to local
production there is a signicant amount of asymmetry whereas in the other markets there
6 This
7A
is implemented in the statistical package STATA using the nlcom command.
benchmark value of 30 percent is used to distinguish a large share from a small share.
18
is no signicant asymmetry.
This is complemented by another measure of market power. This measure is a regression
run specically on items where the presence of one of the main wholesalers is present, namely
Cargill. As mentioned in section 3, this data has been collected from Cargill's country websites. This is an item specic information that will enable us to separate the regressions into
two groups, one where Cargill is present and the other where it isn't. A larger asymmetry for
the group where Cargill is present will point to an inuence of the wholesaler on asymmetry.
The results of these regressions are given in table 5. We see that market power plays a
role in explaining asymmetry. The presence of Cargill signicantly increases the asymmetry
of price transmission whereas when it is absent from a market, the asymmetry term isn't
signicantly dierent from zero.
The presence of one of the main intermediaries as well as the share of local production
exported abroad are both explaining asymmetry, supporting the theory presented above.
6
Conclusion
A model of price transmission from international agricultural prices to producer prices is
presented in order to understand the mechanisms behind price transmission. Due to the geographical dispersion of farmers and economies of scale in wholesaling, agricultural markets
will we characterized by market power on the demand side. The model predicts that this
power of intermediaries buying the products from farmers leads to an asymmetric price transmission when intermediaries have suciently convex marginal cost curves. The asymmetry
is such that there is more price transmission when prices fall.
The results are shown using a Two Stage Least Squares estimator to control for endogeneity problems. This approach has the advantage, as noted by Acharya et al. (2011), of
avoiding some of the disadvantages of many studies in the recent literature that focus on
the time-series properties of the data such that it is sometimes unclear whether rejection
19
of symmetry isn't simply due to specication error. The instruments used are variations in
rainfall, temperature, climate disasters and cloud cover in other geographic regions. The
exclusion of neighbouring countries within the region ensures that these instruments aren't
correlated to the local weather conditions, thus local prices. The results are clear in pointing
out the presence of asymmetry, with the transmission being stronger for price decreases.
Robustness checks using two dierent specications of a Generalized Method of Moments
estimator conrm the presence of an asymmetric price transmission.
The link between market power and asymmetry is then tested. A variable indicating the
presence of intermediaries on specic markets is used to test this link. More specically it
is the presence of Cargill, one of the largest intermediaries in commodity markets, that is
used as a variable and the results show that asymmetry is stronger when Cargill is present.
This result is supported by another regression where the larger the share of local production
exported abroad, the higher the degree of asymmetric price transmission.
This points towards some policy issues, notably the fact that governments should be
aware of the eect of market power of intermediaries and the role they play in inuencing
the price received by farmers, and therefore the gains from trade liberalization in agricultural
markets. Abuse of monopsony power by large intermediaries in agricultural markets can be
particularly harmful in poor countries where farmers often live close to the poverty line.
Future research could be done in the collection of market power data at the item and
country level on a yearly basis. Other private companies than the one looked at here could
also be integrated into such studies. Measures of search costs should also be considered.
20
References
Acharya, R., Kinnucan, H. & Caudill, S. (2011), `Asymmetric farm-retail price transmission
and market power: a new test', Applied Economics 43(30), 47594768.
Arellano, M. & Bond, S. (1991), `Some tests of specication for panel data: Monte carlo
evidence and an application to employment equations', The Review of Economic Studies
58(2), pp. 277297.
Arellano, M. & Bover, O. (1995), `Another look at the instrumental variable estimation of
error-components models', Journal of Econometrics 68(1), 2951.
Blundell, R. & Bond, S. (1998), `Initial conditions and moment restrictions in dynamic panel
data models', Journal of Econometrics 87(1), 115143.
Borenstein, S., Cameron, A. C. & Gilbert, R. J. (1997), `Do gasoline prices respond asymmetrically to crude oil price changes?', The Quarterly Journal of Economics 112(1), 30539.
De Hoyos, R. & Medvedev, D. (2011), `Poverty eects of higher food prices: A global perspective', Review of Development Economics 15(3), 387402.
Galeotti, M., Lanza, A. & Manera, M. (2003), `Rockets and feathers revisited: an international comparison on european gasoline markets', Energy Economics 25(2), 175190.
Gopinath, M., Sheldon, I. M. & Echeverria, R. (2007), `Firm heterogeneity and international
trade: Implications for agricultural and food industries', International Agricultural Trade
Research Consortium, Trade Policy Issues Papers #5 .
Howard, P. (2006), `Consolidation in food and agriculture: Implications for farmers & consumers', The Natural Farmer, publication of Nofa Spring 2006, 1720.
Kinnucan, H. & Zhang, D. (2013), `Perfect farm-retail price transmission',
manuscript .
Levinsohn, J. (1994), `Competition policy and international trade', Working
National Bureau of Economic Research Working Paper Series .
unpublished
paper No. 4972,
MacLaren, D. & Josling, T. (1999), `Competition policy and international agricultural trade',
Working paper #99-7, International Agricultural Trade Research Consortium Working Paper Series .
Mitchell, T., Carter, T., Jones, P., Hulme, M. & New, M. (2003), `A comprehensive set of
high-resolution grids of monthly climate for europe and the globe: the observed record
(1901-2000) and 16 scenarios (2001-2100)', Journal of Climate: submitted .
Mosley, P. & Suleiman, A. (2007), `Aid, agriculture and poverty in developing countries',
Review of Development Economics 11(1), 139158.
Mundlak, Y. & Larson, D. F. (1992), `On the Transmission of World Agricultural Prices',
World Bank Econ Rev 6(3), 399422.
21
Murphy, S. (2006), `Concentrated market power and agricultural trade',
logue Discussion Papers (1).
Ecofair Trade Dia-
Organisation for Economic Co-operation and
Development, Directorate for Financial and Enterprise Aairs Competition CE, Policy
Roundtables .
OECD (2008), `Monopsony and buyer power',
Rogers, R. T. & Sexton, R. J. (1994), `Assessing the importance of oligopsony power in
agricultural markets', American Journal of Agricultural Economics 76(5), pp. 11431150.
Roodman, D. (2006), How to do xtabond2: An introduction to `dierence' and `system'
gmm in stata, Working papers, Center for Global Development.
Sexton, R. J. (1990), `Imperfect competition in agricultural markets and the role of cooperatives: A spatial analysis', American Journal of Agricultural Economics 72(3), pp.
709720.
Shaw, L. H. (1964), `The eect of weather on agricultural output: A look at methodology',
Journal of Farm Economics 46(1), pp. 218230.
Tappata, M. (2009), `Rockets and feathers: Understanding asymmetric pricing',
Journal of Economics 40(4), 673687.
RAND
The OFDA/CRED International Disaster Database (n.d.). EM-DAT: The OFDA/CRED International Disaster Database, www.emdat.be, Université Catholique de Louvain, Brussels
(Belgium).
Tomek, W. G. & Myers, R. J. (1993), `Empirical analysis of agricultural commodity prices:
A viewpoint', Review of Agricultural Economics 15(1), 181202.
Vink, N. & Kirsten, J. (2002), `Pricing behaviour in the south african food and agricultural
sector'. Report commissioned by the National Treasury, Pretoria.
World CIA Factbook (2011).
factbook/elds/2096.html.
https://www.cia.gov/library/publications/the-world-
World Development Indicators (2010). World dataBank, The World Bank Group.
22
Appendix
Appendix 1: Sample of items.
Item
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Almonds, with shell
Anise, badian, fennel, corian.
Apples
Apricots
Arecanuts
Artichokes
Asparagus
Avocados
Bananas
Barley
Beans, dry
Beans, green
Beeswax
Berries Nes
Blueberries
Broad beans, horse beans, dry
Buckwheat
Cabbages and other brassicas
Canary seed
Carobs
Carrots and turnips
Cashew nuts, with shell
Castor oil seed
Cattle meat
Cauliowers and broccoli
Cereals, nes
Cargill
No
No
Yes
No
No
No
No
No
No
Yes
Yes
Yes
No
No
No
No
No
No
No
No
No
No
No
Yes
No
No
Item
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
23
Cherries
Chestnuts
Chick peas
Chicken meat
Chicory roots
Chillies and peppers, dry
Chillies and peppers, green
Citrus fruit, nes
Cloves
Cocoa beans
Coconuts
Coee, green
Cotton lint
Cottonseed
Cow milk, whole, fresh
Cow peas, dry
Cranberries
Cucumbers and gherkins
Currants
Dates
Duck meat
Eggplants (aubergines)
Fibre Crops Nes
Figs
Flax bre and tow
Fruit Fresh Nes
Cargill
No
Yes
No
Yes
No
No
No
No
No
Yes
No
Yes
Yes
Yes
No
No
No
No
No
No
Yes
No
No
No
No
No
Appendix 1: Sample of items (continued).
Item
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
Fruit, tropical fresh nes
Game meat
Garlic
Ginger
Goat meat
Goose and guinea fowl meat
Gooseberries
Grapefruit (inc. pomelos)
Grapes
Groundnuts, with shell
Hazelnuts, with shell
Hemp Tow Waste
Hempseed
Hen eggs, in shell
Hops
Horse meat
Jute
Karite Nuts (Sheanuts)
Kiwi fruit
Leeks, other alliaceous veg
Leguminous vegetables, nes
Lemons and limes
Lentils
Lettuce and chicory
Linseed
Lupins
Maize
Maize, green
Mangoes, mangosteens, guavas
Manila Fibre (Abaca)
Maté
Meat nes
Millet
Mixed grain
Mushrooms and trues
Mustard seed
Natural honey
Cargill
No
No
No
No
Yes
Yes
No
Yes
No
Yes
No
No
No
Yes
Yes
Yes
No
No
No
No
No
No
No
Yes
No
No
Yes
Yes
No
No
No
Yes
Yes
No
No
No
No
Item
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
24
Natural rubber
Nutmeg, mace and cardamoms
Nuts, nes
Oats
Oilseeds, Nes
Okra
Olives
Onions (inc. shallots), green
Onions, dry
Oranges
Other Bastbres
Other bird eggs,in shell
Other melons (inc.cantaloupes)
Palm kernels
Palm oil
Papayas
Peaches and nectarines
Pears
Peas, dry
Peas, green
Pepper (Piper spp.)
Persimmons
Pig meat
Pigeon peas
Pineapples
Pistachios
Plantains
Plums and sloes
Poppy seed
Potatoes
Pulses, nes
Pumpkins, squash and gourds
Pyrethrum,Dried
Quinces
Rabbit meat
Ramie
Rapeseed
Cargill
No
No
Yes
Yes
Yes
No
No
No
No
No
No
No
No
Yes
Yes
No
No
No
Yes
Yes
No
No
Yes
No
No
No
No
No
No
No
No
No
No
No
Yes
No
Yes
Appendix 1: Sample of items (continued).
Item
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
156
157
158
159
160
161
Cargill
Raspberries
Rice, paddy
Roots and Tubers, nes
Rye
Saower seed
Sesame seed
Sheep meat
Silk-worm cocoons, reelable
Sisal
Sorghum
Sour cherries
Soybeans
Spices, nes
Spinach
Stone fruit, nes
Strawberries
Sugar beet
Sugar cane
Sunower seed
Sweet potatoes
Tangerines, mandarins, clem.
Taro (cocoyam)
Walnuts, with shell
Watermelons
Wheat
Wool, greasy
Yams
Yautia (cocoyam)
25
No
Yes
No
Yes
No
No
Yes
No
No
Yes
No
Yes
Yes
No
No
No
Yes
Yes
Yes
No
No
No
No
No
Yes
No
No
No
Appendix 2: Sample of countries.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Albania
Algeria
Argentina
Armenia
Australia
Austria
Azerbaijan
Bangladesh
Barbados
Belarus
Belgium
Belize
Bhutan
Bolivia
Bosnia and Herzegovina
Brazil
Brunei Darussalam
Bulgaria
Burkina Faso
Burundi
Cambodia
Cameroon
Canada
Chile
China
Congo
Costa Rica
Croatia
Cuba
Cyprus
Czech Republic
Denmark
Dominican Republic
Ecuador
Egypt
El Salvador
Eritrea
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
Estonia
Ethiopia
Finland
France
Georgia
Germany
Ghana
Greece
Guinea
Honduras
Hungary
India
Indonesia
Iran, Islamic Republic of
Ireland
Israel
Italy
Jamaica
Japan
Jordan
Kazakhstan
Kenya
Korea, Republic of
Kyrgyzstan
Lao People's Democratic Republic
Latvia
Lebanon
Lithuania
Luxembourg
Madagascar
Malawi
Malaysia
Mali
Mauritius
Mexico
Moldova
Mongolia
26
Appendix 2: Sample of countries (continued).
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
Morocco
Mozambique
Myanmar
Namibia
Nepal
Netherlands
New Zealand
Nicaragua
Niger
Nigeria
Norway
Pakistan
Panama
Paraguay
Peru
Philippines
Poland
Portugal
Romania
Russian Federation
Rwanda
Saudi Arabia
Slovakia
Slovenia
South Africa
Spain
Sri Lanka
Sudan
Sweden
Switzerland
Syrian Arab Republic
Tajikistan
Thailand
Togo
Trinidad and Tobago
Tunisia
Turkey
112
113
114
115
116
117
27
Ukraine
United Kingdom
United States of America
Uruguay
Yemen
Zimbabwe
Table 1: Asymmetry of agricultural price transmission.
OLS
IV
VARIABLES
ln(prod price) ln(prod price)
ln(exp price)
0.4393***
0.9783***
-0.0014
2.2348**
-0.0122***
-0.3747**
40174
0.3285
4978
40174
(0.0083)
price up
(0.0229)
price up * ln(exp price)
(0.0035)
N
R2
Number of clusters
Number of instruments
Kleibergen-Paap rk LM statistic
Hansen J statistic
(0.1205)
(1.0889)
(0.1712)
4978
12
P-val = 0.0000
P-val = 0.4260
Standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1
Clustering by item-country.
28
Table 2: First stage regressions
ln(exp price)
ln(temperature IV)
-0.5035***
(0.0399)
ln(cloud cover IV)
0.2892***
(0.0726)
ln(climate disaster IV)
0.2025***
(0.0107)
ln(rainfall IV)
-0.0499
(0.0501)
ln(temp * price up)
0.0389
(0.0326)
ln(cloud cover * price up)
-0.0981
(0.0683)
ln(climate disaster * price up)
-0.0137
(0.0117)
ln(rainfall * price up)
0.0710
(0.0711)
ln(temp * temp up)
0.0072
(0.0191)
ln(cloud cover * cloud cover up)
0.0375
(0.0320)
ln(climate disaster * clim. disast. up)
0.0077
(0.0098)
ln(rainfall * rainfall up)
-0.0098
(0.0209)
N
40174
2
Partial R
0.0551
F(12, 4977)
46.49
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
29
of the 2SLS.
price up
ln(exp price * price up)
-0.0292*
-0.5175***
(0.0171)
(0.1165)
0.0281
0.3575
(0.0363)
(0.2489)
-0.0369***
-0.0995***
(0.0040)
(0.0263)
-0.0732***
-0.5280***
(0.0221)
(0.1511)
0.0459*
0.3825**
(0.0241)
(0.1656)
0.0323
-0.0461
(0.0571)
(0.4047)
0.1387***
0.8697***
(0.0076)
(0.0520)
0.0481
0.2388
(0.0499)
(0.3395)
0.0415***
0.2282**
(.0138)
(0.0946)
0.0130
0.1905
(.0267)
(0.1892)
0.0159***
0.1159***
(.0048)
(0.0331)
0.0116
0.1050
(0.0146)
(0.0996)
40174
40174
0.0324
0.0284
156.95
132.05
Table 3: GMM regressions.
Lags as IV's Adding economic instruments
lag ln(producer price)
0.8028***
0.8129***
(0.0131)
(0.0129)
price up
0.1976***
0.1890***
(0.0578)
(0.0576)
ln(exp price)
0.1744***
0.1615***
(0.0129)
(0.0127)
price up * ln(exp price)
-0.0193**
-0.0181**
(0.0085)
(0.0085)
year dummies
Yes
Yes
N
28483
28483
AR(1)
0.0000
0.0000
AR(2)
0.1137
0.1205
Hansen J statistic
0.8050
0.7252
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
Table 4: Long and short run transmission.
lags as IV's
Adding economic instruments
short run long run short run
long run
ln(exp price)
0.1744*** 0.8840*** 0.1615***
0.8631***
(0.0129)
(0.0373)
(0.0127)
(0.0381)
price up * ln(exp price) -0.0193** -0.1791** -0.0181**
-0.1470*
(0.0085)
(0.0740)
(0.0085)
(0.0771)
Variable
Table 5: Asymmetry explained by the export importance.
large exp share
small exp share
OLS
2SLS
OLS
2SLS
ln(exp price)
0.6403*** 1.1080*** 0.4666*** 0.6715***
(0.0108)
(0.1908)
(0.0108)
(0.1621)
price up
-0.1747***
2.3992
-0.0416
-0.6772
(0.0558)
(1.5583)
(0.0324)
(1.5970)
price up * ln(exp price)
0.0020
-0.4344*
-0.0069
0.0497
(0.0079)
(0.2376)
(0.0051)
(0.2402)
N
16343
16343
23831
23831
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
30
Table 6: Cargill presence.
Cargill
No Cargill
OLS
2SLS
OLS
2SLS
ln(exp price)
0.5168*** 0.9979*** 0.4709*** 0.9258***
(0.0130)
(0.1196)
(0.0105)
(0.1582)
price up
-0.0180
1.3850
-0.0178
1.7369
(0.0358)
(0.9782)
(0.0332)
(1.5599)
price up * ln(exp price) -0.0143*** -0.2533*
-0.0099*
-0.2886
(0.0055)
(0.1539)
(0.0051)
(0.2437)
N
14692
14692
25482
25482
Standard errors in parentheses
* p < 0.10, ** p < 0.05, *** p < 0.01
Figure 1: Proportion of population in agriculture and wealth.
31
P ∗ ,pf
mc(Q)
P ∗2
w(Q)
P ∗0
P ∗21
pf
p0f
p1f
Q1
Q0 Q2
Q
Figure 2: Price transmission for various international prices.
pf
P∗
Figure 3: Relationship between international price and producer price.
32
Figure 4: Rainfall IV. In parenthesis the export share and the annual standard deviation of
rainfall in millimeters.
33