You are on page 1of 13

The Journal of Economic Asymmetries 23 (2021) e00185

Contents lists available at ScienceDirect

The Journal of Economic Asymmetries


journal homepage: www.elsevier.com/locate/jeca

Asymmetric price responses of the US pork retail prices to farm and


wholesale price shocks: A nonlinear ARDL approach
Dimitrios Panagiotou
Department of Economics, University of Ioannina, Ioannina, Greece

A R T I C L E I N F O A B S T R A C T

JEL classification: The objective of this study is to assess asymmetric price transmission in magnitude and asym-
Q11 metric price transmission in speed, between the farm-retail and wholesale-retail levels of the US
Q13 pork supply chain. In doing so, it employs the recently developed Non-Linear Auto-Regressive
C13
Distributed Lag (NARDL) model. Data utilized are monthly observations for farm, wholesale and
Keywords: retail prices of the US pork supply chain for the period 1990–2018. The empirical findings indicate
Pork
the presence of asymmetry in magnitude for the wholesale-retail pair and the presence of both
NARDL model
asymmetry in speed and asymmetry in magnitude for the farm-retail pair. In the long-run, positive
Price transmission
Asymmetries shocks in the farm prices and in the wholesale prices are transmitted to the retail level with greater
intensity as compared to negative ones.

1. Introduction

Asymmetric price transmission is an increasingly disputed topic in agricultural and food economics. Increases in farm prices are
believed to be transmitted faster at the retail level, whereas negative price shocks at the farm/wholesale level take more time to be
passed on to consumers (Gervais, 2011; Panagiotou and Stavrakoudis, 2017).
Empirical research on vertical price transmission has been undertaken with a variety of statistical tools and econometric techniques.
The simple ordinary least squares and correlation analysis, the smooth transition cointegration model, the linear error correction model
(ECM), the asymmetric/non-linear ECM, the threshold vector ECM, the threshold asymmetric error-correction model (TAECM), the
Markov-switching ECM and the statistical tool of parametric and non-parametric copulas are amongst the most popular tools utilized to
assess the nature of price relationships in the food industries. Most of the findings indicate that price increases at the farm level are
transmitted to the wholesale and to the retail levels faster and/or more fully than price decreases. In the majority of the studies,
increased concentration in downstream markets and input manufacturing has been often pointed out to explain why decreases in
upstream prices are not accompanied by proportional decreases in downstream markets.
In the US hog supply chain, there are studies that have found both symmetric and asymmetric price relationships. More specifically,
Boyd and Brorsen (1988) tested for asymmetry in price adjustments and the speed of price adjustments in the pork marketing channel.
Results revealed symmetric responses. On the other hand, Goodwin and Harper (2000) used a threshold co-integration model to test for
asymmetric adjustment to positive and negative price shocks. Their results revealed important asymmetries. Furthermore, price
adjustment patterns were unidirectional, and information tends to flow from farm, to wholesale, to retail markets.
In more recent studies, results reveal, again, both symmetries and asymmetries in price transmission. Gervais (2011) employed the
smooth transition cointegration method and found evidence of asymmetric long-run transmission in the relationship between prices at

E-mail address: dpanag@uoi.gr.

https://doi.org/10.1016/j.jeca.2020.e00185
Received 13 July 2020; Received in revised form 28 September 2020; Accepted 27 October 2020
1703-4949/© 2020 Published by Elsevier B.V.
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

the farm and at the retail level, along the US pork supply chain. On the other hand, no evidence in favor of short-run asymmetries was
found. Qiu and Goodwin (2012) employed copulas to capture asymmetries along the US pork marketing chain. Their results indicated
that the farm-to-retail price adjustments exhibit a dynamic, time-varying relationship which may reflect the market power of retailers.
The authors indicate that retail prices are more likely to adjust accordingly to a positive price shock at the farm level, than they are to
adjust to negative price shock. This result is consistent with the market power hypothesis as well. On the other hand, the
wholesale-to-retail price adjustments have relatively constant dependence structures, which may reflect that both wholesalers and
retailers have some kind of market power. The study by Emmanoulides and Fousekis (2014) used copulas, but split their sample into two
time periods, to assess the strength and the pattern of price transmission along the US pork. In the first half of the sample (1970–1990),
there was asymmetric co-movement for the pair wholesale-retail. In the second half of the sample (1990–2012), the association between
price changes at the wholesale and the retail level became very weak and symmetric (upper level and lower level of dependence were
equal).
The US pork industry has undergone dramatic changes over the last thirty years (Key and McBride, 2007). Increased concentration in
packing has been accompanied by increased ownership of livestock. Furthermore, the co-existence of investors owned firms along with
agricultural cooperatives has played a significant role in price formation (Panagiotou and Stavrakoudis, 2018).
At the farm level of the US pork supply chain, hog enterprises have grown larger. By 2010, the largest 110 hog farms, with more than
50,000 hogs each, held over 54 percent of the entire US hog inventory (United States Department of Agriculture–Economic Research
Service, 2017). Most pork production is under vertical control, either through vertical integration or production contracts (Norwood and
Lusk, 2018). Alternative marketing arrangements including packer ownership, is estimated at 89% of the hog volume (Greene, 2015).
Hence, hog producers might have some leverage on the distribution of added-value along the pork supply chain.
The wholesale level of the pork marketing chain has undergone rapid concentration in the last thirty years. In 1982, the top four hog
packers (CR4) controlled 36% of the market. In 2006, their share had increased to 62% percent. By 2010, the CR4 has risen to 67%. More
specifically, Smithfield with 31%, Tyson Foods with 17%, Swift& Co. with 11% and Cargill with 8% control two thirds of the market
(U.S. Government Accountability Office (GAO), 2009). In addition to pork packing, the top four pork packers are also involved in other
areas of agribusiness. As a consequence, agents at the wholesale hold significant market power.
At the retail level of the US pork supply chain we meet also high levels of concentration. Retail grocers like Kroger and Walmart, food
service providers like Compass Group PLC, and restaurants/fast food chains like McDonald’s and Burger King are players with important
name recognition and might have a significant degree of leverage. Evidence of this might be reflected in Fig. 1, where retail prices
exhibit a constantly increasing trend up to the year of 2015. For the period 2015–2020 retail prices have remained (on average)
constant. On the other hand, wholesale and especially farm prices, have ups but also significant downs. Between the time period
2014–2015, both farm and wholesale prices drop significantly and for the remaining period they both remain (on average) at the price
level of 2015.
In the light of the preceding, and under the increasingly changing conditions in the US pork sector, the present study investigates to
what extent and degree, price shocks at the farm and at the wholesale levels of the US pork supply chain are transmitted to the retail
level.
In doing so, it employs the recently developed nonlinear ARDL (NARDL) cointegration approach by Shin et al. (2014, pp. 281–314)

Fig. 1. The natural logarithms of farm, wholesale and retail prices of the US pork supply chain.

2
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

and monthly data for the period 1990–2018. The NARDL approach exhibits significant advantages over the existing statistical tools and
econometric models that have been used so far. These advantages are the detection of cointegrating relationships in small samples
efficiently, the joint modelling of asymmetries and cointegration dynamics, the joint analysis of non-stationarities and non-linearities,
and the identification of asymmetries both in the long-run and in the short-run. Furthermore, it does not require the regressors to be
integrated of the same order, allowing for the inclusion of both I (0) and I (1) (but not I (2)) time series processes in the long-run
equilibrium relationship.
To the best of our knowledge, there has been only a handful of works on asymmetric price transmission in the food markets using the
NARDL model: Fousekis et al. (2016) investigate vertical price transmission along the US beef supply chain, Fousekis and Trachanas
(2016) examine spatial price linkages in the skimmed milk powder markets USA, EU and Oceania, Rezitis (2018) analyzes vertical price
transmission between farm and retail markets for a variety of dairy products in Finland, and Bronnmann and Bittmann (2019, p.
103513) investigate vertical price transmission between import and retail markets for cod and herring in Germany.
The objective of this study is to assess asymmetric price transmission in magnitude and asymmetric price transmission in speed,
between the farm-retail and wholesale-retail levels of the US pork supply chain, with the recently developed NARDL model.
In what follows, section 2 provides the theoretical model. Section 3 presents the data and Section 4 provides the empirical models
and testing. Section 4 offers the empirical results and Section 5 presents the conclusions.

2. The NARDL cointegration methodology

The recent nonlinear autoregressive distributed lag (NARDL) model proposed by Shin et al. (2014, pp. 281–314) is employed in the
present study for detecting both long- and short-run (a)symmetric price responses of the US pork retail prices when farm and wholesale
prices change.
Let’s assume that we have two time series variables: yt is the dependent variable and xt is the independent variable. The NARDL
approach accounts for asymmetry by decomposing the independent variable xt into its positive and negative partial sums (Shin et al.,
2014, pp. 281–314) as follows:

xt ¼ x0 þ xþ 
t þ xt (1)

where

X
t X
t

t ¼ Δxþ
i ¼ maxΔðxi ; 0Þ (2)
i¼1 i¼1

and

X
t X
t
x
t ¼ Δx
i ¼ minΔðxi ; 0Þ (3)
i¼1 i¼1

Accordingly, the long-run equilibrium relationship is expressed as:

yt ¼ βþ xþ  
t þ β xt þ ut (4)

where βþ and β are the asymmetric long-run parameters associated with positive and negative changes in the independent variable xt ,
respectively.
Shin et al. (2014) demonstrated that by combining equation (4) with the ARDL (p,q) model (Pesaran and Shin, 1999) we obtain the
NARDL (p,q) model:

X
p1
Δyt ¼ α0 þ ρ yt1 þ θþ xþ  
t1 þ θ xt1 þ γ j Δytj
j¼1
(5)
X
q1 X
q1
þ π þj Δxþtj þ π j Δxtj þ et
j¼0 j¼0

where θþ ¼ ρ=βþ and θ ¼  ρ=β . The letters p and q capture the lag orders for the dependent variable (yt ) and the independent
variable (xt ) in the part of the distributed lag, respectively.
In order to proceed with the empirical implementation of the NARDL model it is necessary to follow the next steps. First of all, unit
root tests must be undertaken to ensure that the variables used in the empirical model are not integrated of order 2, namely I (2).
Secondly, the NARDL model of equation (5) is estimated using the standard OLS technique. The third step is to test and verify (or not) for
the existence of an asymmetric cointegrating relationship between the levels of the series yt ; xtþ and xt . This is undertaken following two
proposed approaches. In the first one is by Shin et al. (2014, pp. 281–314), the null hypothesis of no cointegration (ρ ¼ θþ ¼ θ ¼ 0) is
tested using the FPSS statistic. In the second approach, the null hypothesis of no-cointegration (ρ ¼ 0 against ρ < 0) is tested using the
WPSS statistic (Pesaran et al., 2001). In the fourth step, we employ Wald tests to examine for long-run and short-run symmetries. More
specifically, for long-run symmetry the null hypothesis to be tested is whether βþ ¼ β . For short-run symmetry, the null hypothesis can

3
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Table 1
Descriptive statistics of data variables (raw data are in cents/pound).
Variable Mean Max Min Std. Dev.

Pf 86.519 166.500 29.300 18.971


Pw 126.121 223.100 81.100 23.359
Pr 286.466 421.500 199.300 60.184
lnPf 4.437 5.115 3.778 0.223
lnPw 4.821 5.407 4.396 0.176
lnPr 5.636 6.044 5.295 0.206

take either one of the following two forms: (i) the pairwise (strong-form) symmetry which tests for π þ 
j ¼ π j for all j ¼ 1; :::; q 1; or (ii)
Pq1 þ Pq1 
the additive (weak-form) symmetry which tests for j¼0 π j ¼ j¼0 π j . Lastly, and provided that there is asymmetry in the long-run or
in the short-run or in both, the asymmetric cumulative dynamic multiplier effects, associated with a unit change in xtþ and in xt , are
derived. The positive and negative dynamic multipliers are calculated as:

Xh
∂ytþj Xh
∂ytþj

h ¼ and m
h ¼ ; h ¼ 0; 1; 2::::: (6)
j¼0
∂ xþ
t j¼0
∂xt
þ 
In equation (6), as h → ∞, then mþ 
h → β and mh → β
Based on the estimated dynamic multipliers, the paths and the length of adjustments, from the initial equilibrium to a new equi-
librium between the variables of the model, can be observed. The disequilibrium is caused by a positive or a negative shock in prices. In
other words, the dynamic multipliers provide the evolution of the price at the retail level of the US pork supply chain towards a new
equilibrium due to a shock to a price at farm and the wholesale levels of the marketing chain. Hence, mþ 
h and mh provide the reader with
useful information regarding the long- and short-run asymmetric patterns of the price pairs farm-retail and wholesale-retail.

3. Data

The data for the empirical analysis come from the Economic Research Service of the United States Department of Agriculture
(ERSUSDA). Data refer to the period 1990:2 to 2018:12. These are monthly prices expressed in cents per pound (retail weight equiv-
alent) at the farm, the wholesale and the retail level of the U.S. pork industry. Table 1 presents the descriptive statistics of the raw prices
for each level of the US pork supply chain as well as their respective natural logarithms.Where Pf ¼ farm price, Pw ¼ wholesale price, Pr
¼ retail price and lnPf , lnPw , lnPr are the natural logarithms of the farm, the wholesale and the retail price, respectively.
Fig. 1 presents the natural logarithms of the prices at the three levels of the US pork supply chain. Retail and wholesale price series
exhibit upward trends. Farm price series exhibit a more constant trend with a quite sharp decrease in the late 1990s.
Fig. 2 presents the changes in the first difference in the natural logarithms of the prices. Accordingly, Fig. 2 presents the percentage
changes in the raw data of farm, wholesale and retail prices of the US pork supply chain. Prices at the farm and at the wholesale level are
much more volatile than those at the retail level.
Table 2 presents the results of the Kwiatkowski et al. (1992) (KPSS) test. The test values suggest that the logarithmic prices, in each
one of the three different market levels - farm, wholesale and retail - contain unit roots. However, all time series when differenced
(namely the price log-returns), are stationary. Table 3 presents the p-values from the Dickey and Fuller (1979) and the Phillips and
Perron (1988) unit root tests (the null hypothesis is the existence of a unit root). Table 4 presents the Zivot and Andrews (2002) unit root
test with one structural break. The results suggest that for the case of the (logarithm) farm prices the null hypothesis of unit root can be
rejected at the 1% level, for both the Dickey-Fuller and the Phillips-Perron test and under both cases, namely with constant (C) and with
constant and linear trend (C/T). For the case of the (logarithm) wholesale prices, the null hypothesis of unit root is (is not) rejected at the
1% level, for both the Dickey-Fuller and the Phillips-Perron test, and under the case of a constant (constant and linear trend) term. For
the case of the (logarithm) retail prices the null hypothesis of unit root is rejected at the 1% level, for both the Dickey-Fuller and the
Phillips-Perron test and under both cases, namely with constant and with constant and linear trend. Lastly, all three time series are
stationary when differenced.
In the empirical investigation of vertical price transmission, the identification of causal markets is of great importance. Causal
markets are the markets at which prices are established. In a number of studies, the direction of causality is defined ad hoc based on
certain characteristics of the market. Usually, it is assumed that prices are established at the farm level and it flows forward to the
wholesale and then to the retail level (Emmanoulides and Fousekis, 2014; Gervais, 2011). On the other hand, some studies have used a
causality or exogeneity test in order to identify the casual markets. The present work, in order to determine at which level of the supply
chain the prices are established, utilizes the leveraged bootstrap simulation test of causality proposed by Hacker & Hatemi-J, 2012. The
later produces critical values that are not sensitive to non-normal errors and to time-varying volatility.1 The test has been applied on the
two pairs of prices: farm-wholesale (lnPf - lnPw ) and wholesale-retail (lnPw - lnPr ). Table 5 presents the results. For the farm-wholesale
pair, and at the 1% level of significance or less, the causal order flows uni-directionally, from the farm level to the wholesale level. For

1
Due to the fact that our variables are I(1), the Wald statistic employed to determine Granger causality (Granger, 1969, 1988) is non-standard.

4
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Fig. 2. Changes (first difference) in the natural logarithms of farm, wholesale and retail prices along the US pork supply chain.

Table 2
KPSS tests.
Levels and Returns Test value

lnPf 0.995***
lnPw 0.887***
lnPr 0.778***
ΔlnPf 0.071
ΔlnPw 0.058
ΔlnPr 0.056

Note: The critical values for the KPSStests are 0.739,


0.463, and 0.347 at the 1, the 5, and the 10 percent level
of significance, respectively.(***, **, *): It denotes the
rejection of thenull hypothesis (stationarity) in favor of
the alternative hypothesis (unit root), at the 1, 5, and 10
percent level, respectively.

Table 3
p-values of the DF and Phillips-Perron unit root tests (null hypothesis is the existence of a unit root).
Dickey-Fuller Phillips-Perron

Variable C C=T C C=T


lnPf 0 (lag ¼ 1) 0.001 (lag ¼ 1) 0.001 (lag ¼ 6) 0.002 (lag ¼ 6)
lnPw 0.009 (lag ¼ 2) 0.002 (lag ¼ 3) 0.005 (lag ¼ 6) 0.005 (lag ¼ 4)
lnPr 0.009 (lag ¼ 1) 0.001 (lag ¼ 1) 0.006 (lag ¼ 1) 0.004 (lag ¼ 3)
ΔlnPf 0 (lag ¼ 1)*** 0 (lag ¼ 1)*** 0 (lag ¼ 16)*** 0 (lag ¼ 16)***
ΔlnPw 0 (lag ¼ 1)*** 0 (lag ¼ 1)*** 0 (lag ¼ 14)*** 0 (lag ¼ 14)***
ΔlnPr 0 (lag ¼ 0)*** 0 (lag ¼ 0)*** 0 (lag ¼ 5)*** 0 (lag ¼ 5)***

Note: Δ is the first difference operator.The optimal lag structure of the Dickey-Fuller test is chosen based on the Schwarz Information Criterion and is
displayed in parentheses. The optimal lag structure of the Phillips–Perron test is chosen based on the Newey–West bandwidth with Bartlett weights and
is displayed in parentheses. The estimations and tests were conducted using EViews 10. (***): rejection of the null hypothesis of a unit root at 1% level
or less.

the pair wholesale-retail, and at the 1% level of significance or less, causality turns out to be uni-directional as well, namely, from the
wholesale to the retail level. Hence, the results are in agreement with the majority of previous empirical evidence, according to which,
meat prices are more likely to be established at the upstream level of the supply chain and to flow to the downstream levels of the US

5
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Table 4
Zivot-Andrews (ZA) unit root tests with one structural break.
Model A Model B Model C

Variable ZA test statistic TB ZA test statistic TB ZA test statistic TB


lnPf 4.660 (12) 2009:08 4.170 (12) 2013:07 4.834 (12) 2009:11
lnPw 5.601 (3) 2009:11 4.735 (3) 1998:11 5.594 (3) 2009:11
lnPr 4.194 (1) 1991:97 4.343 (1) 1991:11 4.192 (1) 1995:06
ΔlnPf 5.497 (11)*** 1998:11 5.212 (11)*** 2010:06 5.523 (11)** 1998:11
ΔlnPw 11.181 (2)*** 2013:06 15.332 (1)*** 1990:5 11.260 (2)*** 2016:06
ΔlnPr 14.784 (0)*** 1990:06 14.719 (0)*** 1990:08 14.842 (0)*** 1990:05

Note: Δ is the first difference operator.TB denotes the time of break. Model A allows a break at an unknown point in the intercept; Model B allows a
break at an unknown point in the linear trend; and Model C allows a break at an unknown point in both. The optimal lag structure of the Zivot and
Andrews (2002) test is chosen based on the AIC Information Criterion and is displayed in parentheses.The critical values were obtained from Zivot and
Andrews (1992). The estimations and tests were conducted using R 3.6.1. (***), (**): rejection of the null hypothesis of a unit root at 1% and 5% level,
respectively.

Table 5
Causality tests by Hacker & Hatemi-J, 2012.
Null hypothesis Test value Bootstrap critical value at 1%

lnPf does not cause lnPw 33.873*** (6) 12.325


lnPw does not cause lnPf 1.633 (6) 12.728
lnPw does not cause lnPr 97.860*** (6) 13.672
lnPr does not cause dlnPw 10.454 (6) 13.987

Note: Optimal lags are in parentheses. The maximum lag was set to 12. Thebootstrap simulations for the causality tests
were conducted using 10,000 simulations.(***): It denotes rejection of the null hypothesis at 1% level or less.

meat (pork and beef) supply chains (Goodwin and Harper, 2000; Goodwin and Holt, 1999).

4. Empirical models and testing

Based on the findings of the causality tests, price is transmitted from farm level to the wholesale and from there to the retail level. The
present study wants to investigate how price shocks at the farm level and at the wholesale are transmitted to the retail level. Hence, the
following general form NARDL models will be tested and the appropriate models will be picked and estimated:
Model 7 represents the farm to retail relationship under long-run and short-run asymmetric price transmission.

X
p1
ΔlnPrt ¼ α0 þ ρ lnPrt1 þ θþ lnPft1
þ
þ θ lnPft1

þ αj ΔlnPrtj
j¼1
(7)
X
q1 X
q1
þ π þj ΔlnPftj
þ
þ π j ΔlnPftj

þ et
j¼0 j¼0

Model 8 captures the farm to retail price transmission relationship when short-run symmetry is imposed.

X
p1
ΔlnPrt ¼ α0 þ ρ lnPrt1 þ θþ lnPft1
þ
þ θ lnPft1

þ αj ΔlnPrtj
j¼1
(8)
X
q1
þ π j ΔlnPftj þ et
j¼0

Model 9 represents the wholesale to retail relationship under long-run and short-run asymmetric price transmission.

X
p1
ΔlnPrt ¼ α0 þ ρ lnPrt1 þ θþ lnPwþ  
t1 þ θ lnPwt1 þ αj ΔlnPwtj
j¼1
(9)
X
q1 X
q1
þ π þj ΔlnPwþtj þ π j ΔlnPwtj þ et
j¼0 j¼0

Model 10 represents the wholesale to retail price transmission relationship when short-run symmetry is imposed.

6
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Table 6
Bounds testing for asymmetric cointegration.
Farm to Retail Wholesale to Retail Wholesale to Retail

SR&LR Asymmetry SR&LR Asymmetry SR symmetry


(Model 7) (Model 9) (Model 10)
Statistic Test statistic Test statistic Test statistic
FPSS 7.539*** 7.849*** 7.822***
WPSS 23.917*** 21.467*** 24.465***

X
p1
ΔlnPrt ¼ α0 þ ρ lnPrt1 þ θþ lnPwþ  
t1 þ θ lnPwt1 þ αj ΔlnPwtj
j¼1
(10)
X
q1
þ π j ΔlnPwtj þ et
j¼0

The variables lnPf þ , lnPf  , lnPwþ , and lnPw , are partial sums of positive and negative changes in lnPf and lnPw, respectively. The lag
order has been determined following the general to specific approach (Fousekis et al., 2016; Greenwood-Nimmo and Shin, 2013; Shin
et al., 2014, pp. 281–314). More specifically, the preferred specification in each case has been selected by starting with max q ¼ max p ¼
12. Subsequently, we drop all the insignificant regressors with a 5% unidirectional decision rule. Including insignificant lags is likely to
lead to inaccuracies in the estimation which may lead to noise into the dynamic multipliers.
Table 6 presents the test results for asymmetric cointegration. In both models 7 and 9, the null hypothesis of no cointegration is
rejected at the 1% or 5% levels of significance. The same holds true for the wholesale-retail pair when short-run is imposed (Model 10).2
Note: Following Shin et al. (2014) and Fousekis et al. (2016), we have adopted the conservative approach to the choice of critical
values by employing k ¼ 1. For the 1% level of significance, the pair of critical values (bounds) for the FPSS and the WPSS statistics are
6.3–7.5 and 14.1 to 15.6, respectively. For the 5% level of significance, the pair of bounds for the FPSS and the WPSS statistics are 4.8–5.8
and 9.8 to 11.5, respectively. The critical values have been obtained from Pesaran et al. (2001). The estimations and tests were con-
ducted using a program code written in E-views retrieved from Matthew Greenwood-Nimmo’s web page. Accordingly the estimations
and tests were conducted using EViews 10. (***): rejection of the null hypothesis of no cointegration at th 1% level of significance.
Table 7 presents the test results for long- and short-run symmetry. The Wald test strongly rejects the null hypothesis of long-run
symmetry for the price pairs farm-retail (lnPf - lnPr) and wholesale-retail (lnPw - lnPr). On the other hand, the Wald test does not
reject the null hypothesis of short-run symmetry in the wholesale-retail pair. It does reject though the null hypothesis of short-run
symmetry for the pair farm-retail. Accordingly, one may conclude that asymmetric price transmission in magnitude is present in the
relationships between the farm-retail as well as the wholesale-retail levels of the pork supply chain in the US. On the other hand,
asymmetric price transmission in speed is rejected in the price relationship between the wholesale-retail levels of the US the pork
marketing chain. Lastly, asymmetric price transmission in speed can not be rejected for the farm-retail price pair.
The estimations and tests were conducted using a program code written in E-views retrieved from Matthew Greenwood-Nimmo’s
webpage.
Based on the test results, the farm-to-retail price relationship will be estimated via model 7, namely under long-run and short-run
asymmetries. On the other hand, the wholesale-to-retail price relationship was tested and short-run symmetry could not be rejected.
Hence, model 10 will used for the estimation of the wholesale-retail price relationship.

5. NARDL estimation results

Table 8 presents the NARDL results for the pair farm-retail under long-run and short-run asymmetries. Table 9 presents the NARDL
results for the pair wholesale-retail with short-run symmetry imposed.
For the price pair farm-retail, all the estimated coefficients are statistically significant. The estimated value of the long-run coefficient
βþ equals 0.481 while that of the coefficient β is equal to 0.404. Accordingly, a 1% increase (decrease) in the farm price leads to a
0:481% (0:404%) increase (decrease) in the retail price of the US pork supply. Hence, in the long-run, positive shocks in the farm prices
are transmitted to the retail level with greater intensity compared to negative ones. The magnitude of the transmission elasticity of
positive price shocks at the farm level is 19% higher than that of negative price shocks.
For the wholesale-retail price pair, all the estimated coefficients are statistically significant as well. The estimated value of the long-
run coefficient βþ equals 0.488 and that of the coefficient β is equal to 0.409. Accordingly, a 1% increase (decrease) in the farm price
leads to a 0:488% (0:409%) increase (decrease) in the retail price of the US pork supply. The magnitude of the transmission elasticity of
positive price shocks at the farm level is 19% higher than that of negative price shocks. Hence, in the long-run, positive shocks in the
farm prices are transmitted to the retail level with greater intensity compared to negative ones. As one might conclude, the estimated
values of the asymmetric long-run price transmission elasticities are very similar for the two price pairs examined here.
Fig. 3 displays the results of the cumulative sum (CUSUM) tests on the selected NARDL models. There is no evidence of statistical

2
We also followed Shin et al. (2014, pp. 281–314), where we adopted the approach to the choice of critical values by employing k ¼ 1 when
testing for the null.

7
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Table 7
Wald test statistics for long-run and short-run symmetry.
Farm-Retail Wholesale-Retail Wholesale-Retail

(SR&LR Asymmetry) (SR&LR Asymmetry) (SR symmetry)


(Model 7) (Model 9) (Model 10)
Statistic Test statistic Test statistic Test statistic
WLR 61.009 (0.000) 54.895 (0.000) 89.115 (0.000)
WSR 4.766 (0.029) 1.225 (0.269) - ()
þ 
b
θ b
θ
Note: WLR refers to the Wald test of long-run symmetry with the null hypothesis defined as:  ¼  WSR denotes the Wald test for the null of
ρ
b ρ
b
P þ
Pq1 
the additive short-run symmetry defined as: q1 π
j¼0 j ¼ π
j¼0 j Numbers in parentheses are p-values.

instability of the parameters.


For the farm-retail price pair, the findings of the present study are partly in agreement with the results obtained by Gervais (2011).
Both studies found evidence of asymmetries in price transmission in the long-run. On the other hand, the study by Gervais (2011) found
no evidence of short-run asymmetric price transmission whereas this work estimates the farm-retail price relationship under short-run
asymmetries. For the wholesale-retail pair, Emmanoulides and Fousekis (2014) concluded that there is asymmetric price co-movement
for the first part of their sample (1970–1990) and no evidence of co-movement for the second part of their sample (1990–2012). The
present work concluded asymmetric price relationship in the long-run but symmetric price relationship in the short-run for the
wholesale-retail pair. Lastly, Rezitis and Tsionas (2019) employed a multivariate panel error correction model (PVECM) to investigate
asymmetric price transmission among the farm, processor, and retail sectors of the European food supply chain for the 2005–2016
period. The authors found that in both the long- and short-run, retail prices respond more strongly to processor price increases than
decreases and the same occurs for processor prices due to farm price changes. Hence, their findings demonstrate the presence of positive
asymmetric price transmission in the European food supply chain. The present study arrived at the same conclusion regarding the
pattern of price transmission but for the case of the US pork supply chain.
Fig. 4 presents the dynamic multipliers for the price transmission from the farm to the retail level. The dynamic multipliers allow us
to trace out the evolution of a price at a given level of the supply chain following a shock to a price at another level of it, providing in this
way a picture of the path to the new equilibrium. For the farm-retail pair, we observe that retail prices do not respond at the same rate
(they are above the zero line) to farm price increases and decreases, both in the short-run and in the long-run. The behavior of the
dynamic multiplier is consistent with short-run asymmetry and long-run asymmetry. Equilibrium is corrected after a few periods of time.
Fig. 5 presents the dynamic multipliers for the price transmission from the wholesale to the retail level. We observe that in the very
short-run, retail prices respond at almost the same rate to positive and negative shocks to wholesale prices (dashed green line very close

Table 8
NARDL results for farm-retail with LR&SR asymmetry (Model 7).
Variable Coefficient Std. errors

Constant 0.313*** 0.088


lnPrt1 0.059*** 0.016
þ 0.028*** 0.008
lnPft1

lnPft1 0.023*** 0.007
ΔlnPrt4 0.117** 0.046
ΔlnPrt7 0.112** 0.044
ΔlnPrt12 0.116*** 0.044
ΔlnPftþ 0.021** 0.010
þ
ΔlnPft1 0.062*** 0.011
þ
ΔlnPft2 0.045*** 0.010
þ
ΔlnPft3 0.023** 0.011
þ
ΔlnPft4 0.021** 0.009

ΔlnPft1 0.025** 0.011

ΔlnPft2 0.037*** 0.011

ΔlnPft3 0.043*** 0.012
Asymmetric long-run price transmission elasticities:
þ
b
θ 0.481*** 0.097
βþ ¼ 
ρ
b

b
θ 0.404*** 0.105
β ¼ 
ρ
b
Statistics:
R2 0.433
Serial Correlation LM test (Breusch-Godfrey’s F-stat.) 0.326 (0.984)
Heteroskadisticity (White’s F-stat.) 0.931 (0.663)

Note: Numbers in parentheses are p-values. The estimations and tests were conducted using a program code written in E-views
retrieved from Matthew Greenwood-Nimmo’s webpage. (***), (**): rejection of the null hypothesis of a unit root at 1% and
5% level, respectively.

8
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Table 9
NARDL results for wholesale-retail with SR symmetry imposed (Model 10).
Variable Coefficient Std. errors

Constant 0.345*** 0.094


lnPrt1 0.064*** 0.018
lnPwþt1 0.026*** 0.008
lnPwt1 0.34** 0.108
ΔlnPrt7 0.108** 0.045
ΔlnPrt12 0.111** 0.044
ΔlnPwt 0.031*** 0.012
ΔlnPwt1 0.069*** 0.014
ΔlnPwt2 0.072*** 0.013
ΔlnPwt3 0.051*** 0.013

Asymmetric long-run price transmission elasticities:


þ
b
θ 0.488*** 0.087
βþ ¼ 
ρ
b

b
θ 0.409*** 0.094
β ¼ 
ρ
b

Statistics:
R2 0.382
Serial Correlation LM test (Breusch-Godfrey’s F-stat.) 0.787 (0.664)
Heteroskadisticity (White’s F-stat.) 1.039 (0.409)

Note: Numbers in parentheses are p-values. The estimations and tests were conducted using a program code written in E-views
retrieved from Matthew Greenwood-Nimmo’s webpage. (***), (**): rejection of the null hypothesis of a unit root at 1% and
5% level, respectively.

to zero). The magnitude, however, of adjustment is larger for positive shocks with the long-run equilibrium correction being achieved
quite fast. The long-run effect, as portrayed by the (green dashed) asymmetry line, indicates that a price increase at the wholesale level is
larger than that of a price decrease. Accordingly, the effect of a positive shock dominates that of a negative one in the long-run but not in
the very short-run. The behavior of dynamic multipliers is consistent with short-run symmetry and long-run asymmetry.
The aforementioned findings suggest that the effect of a positive shock dominates that of a negative one in both cases (farm-to-retail
and wholesale-to-retail) and there are also significant differences the adjustment between short run and long run. More specifically, the
magnitude of adjustment in the long run is larger for positive shocks indicating that the effects at the retail level of price increases at the
upstream levels are larger than that of price decreases. Given the estimated patterns of price transmission and price adjustments, final
consumers of pork are more likely to feel an increase of prices in the upstream market level (retail) rather than a decrease. Regarding the
actors along the US pork supply marketing chain, retailers enjoy an advantage over wholesalers (pork packers) as well as farmers
(primary producers): their gross margin is more likely to remain the same following a price increase at the wholesale level and it is more
likely to expand following a price decrease at the farm level. Accordingly, retailers, ceteris paribus, are more likely not to experience
losses in their gross profits.
Hence, in the light of the empirical findings of this work, one may conclude that retailers enjoy an advantage over farmers as well as
on wholesalers along the US pork marketing chain. Given that processing and retailing costs remain the same, retailers’ margin is more
likely to remain the same following a price increase at the farm and/or at the wholesale level. Concurrently, their margin is more likely
to widen following a price decrease at the farm/wholesale level. As Fousekis et al. (2016) points out, all of the above can be attributed to
the fact that contemporary food supply chains at the retail level are gaining bargaining power at the expense of processors and primary
producers.
According to the results by Hahn et al. (2004), due to the fact that prices in the hog/pork sector adjust more quickly upward than
downward, actual prices tend to be higher than their full-adjustment values. For the case of the hog sector, farm prices average around
1% higher under asymmetric adjustment than it would be with complete adjustment. The findings of the present study also reveal
asymmetries in the speed of price transmission for the farm-retail pair. Hence, final consumers of pork, when there is a price shock at the
farm level, are more likely to experience an increase of prices at the retail level rather than a decrease. The latter is related to the fact that
meat consumers (at retail level) have become very conscientious about their diet and this has affected their willingness to pay for
different meat types, especially pork and beef (Fogel and Grotte, 2011).

6. Conclusions

The US pork industry has undergone dramatic changes over the last thirty years. Increased concentration in packing has been
accompanied by increased ownership of livestock. Vertical linkages in the pork industry have also changed dramatically in the last
decade: an increased numbers of marketing contracts link packers with hog producers. Furthermore, with the Livestock Mandatory
Reporting Act being in effect for more than fifteen years, the flow and the transparency of information has increased along the US pork
supply chain.
Asymmetric price transmission along food supply chains or across spatial agricultural and food markets have captured for a long

9
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Fig. 3. Cumulative sum (CUSUM) tests on NARDL model 7 (panel A) and NARDL model 10 (panel B).

period of time the attention of researchers because of their implications for the welfare of the primary producers, the final consumers
and the economic efficiency of the markets.
Against this background, the objective of the present study has been to investigate asymmetries in magnitude and in speed in the
transmission of prices along the farm-retail and wholesale-retail levels of the US pork marketing chain. In doing so, the study applies the
NARDL technique that can capture both long- and short-run non-linearities. CUSUM tests Brown et al. (1975) on the selected NARDL

10
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Fig. 4. Dynamic multipliers for the farm-retail pair.

models verify that there is no evidence of statistical instability of the parameters. The empirical findings indicate the presence of
asymmetry in magnitude for the wholesale-retail pair and the presence of both asymmetry in speed and asymmetry in magnitude for the
farm-retail pair. In the long-run, positive shocks in farm prices as well as in wholesale prices are transmitted to the retail level with
greater intensity as compared to negative ones. The magnitude of the long-run price transmission elasticities is very similar for the two
pairs examined in this study.
The dynamic multipliers reveal that short-run and long-run equilibrium is corrected after a few time periods. Given the fact that this
work finds asymmetries in the speed of transmission for the farm-retail pair and equilibrium is corrected relatively fast, hog prices might
be higher than they would under symmetry in the speed of transmission.
The empirical results indicate asymmetries in the speed and in the magnitude of price transmission between the farm-retail and
wholesale-retail levels of the US pork supply chain. Under asymmetry in speed, positive and negative shocks to prices at a given market
level eventually pass to other market levels, even though shocks of different signs have different adjustment paths to the equilibrium.
Asymmetry in magnitude in the present study suggests that prices at the retail level of the supply chain do not return to the same level
after equivalent positive and negative shocks to prices to the upstream levels of the marketing supply chain. As a result, long-run margins
and price ratios are affected. In addition, the empirical findings reveal that the long-run transmission elasticity for positive price shocks
is higher than that for negative price shocks. The latter indicates that retailers’ gross margin is more likely to remain the same following
a price increase at the farm and/or at the wholesale level and it is more likely to expand following a price decrease at the farm and/or at
the wholesale level. The empirical findings of the present study, when viewed in the light of the potential leverage of various actors in
the US pork sector as well as the theoretical works of Azzam (1999) and Xia (2009), might raise concerns about the efficiency of the pork
supply chain in the USA and point to the market power as the possible culprit. More specifically, if market power is present, this type of
long-run asymmetric price transmission is consistent with a price transmission function which is convex in prices. Azzam (1999)
demonstrated that when the power lies with the sellers, strict concavity of the aggregate demand function implies strict convexity of the
price transmission function. Xia (2009), proved that buyer power is more important relative to seller power in determining the concavity
or convexity of a price transmission function and, the extent of the asymmetric price transmission in magnitude. According to Xia’s
(2009) results a strictly concave (convex) aggregate supply function is associated with a strictly convex (concave) price transmission

11
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Fig. 5. Dynamic multipliers for the wholesale-retail pair.

function. Lastly, the findings of this article are in line with the contemporaneous perceptions about food marketing chains which suggest
that retailers are gaining more bargaining power at the expense of wholesalers (Crespi et al., 2012). In the light of all the aforementioned
evidence, US consumers of pork are more likely to feel an increase in prices at the retail level of the pork marketing chain rather than a
decrease.
In the light of the preceding, final consumers of pork, when there is a price shock at the farm level or at the wholesale level are more
likely to experience an increase of prices at the retail level rather than a decrease.

Declaration of competing interest

On behalf of all authors, the corresponding author states that there is no conflict of interest.

References

Azzam, A. M. (1999). Asymmetry and rigidity in farm-retail price transmission. American Journal of Agricultural Economics, 81(3), 525–533.
Boyd, M. S., & Brorsen, B. W. (1988). Price asymmetry in the US pork marketing channel. North Central Journal of Agricultural Economics, 10(1), 103–109.
Bronnmann, J., & Bittmann, T. (2019). Asymmetric adjustment of retail cod and herring prices in Germany: A NARDL approach. Marine Policy.
Brown, R. L., Durbin, J., & Evans, J. M. (1975). Techniques for testing the constancy of regression relationships over time. Journal of the Royal Statistical Society: Series B,
37(2), 149–163.
Crespi, J. M., Saitone, T. L., & Sexton, R. J. (2012). Competition in us farm product markets: Do long-run incentives trump short-run market power? Applied Economic
Perspectives and Policy, 34(4), 669–695.
Dickey, D. A., & Fuller, W. A. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association,
74(366a), 427–431.

12
D. Panagiotou The Journal of Economic Asymmetries 23 (2021) e00185

Emmanoulides, C., & Fousekis, P. (2014). Vertical price transmission in the US pork industry: Evidence from copula models. Agricultural Economics Review,
15(389–2016-23505), 86–97.
Fogel, R. W., & Grotte, N. (2011). Major findings from the changing body: Health, nutrition, and human development in the western world since 1700. The Journal of
Economic Asymmetries, 8(2), 1–9.
Fousekis, P., Katrakilidis, C., & Trachanas, E. (2016). Vertical price transmission in the US beef sector: Evidence from the nonlinear ARDL model. Economic Modelling,
52, 499–506.
Fousekis, P., & Trachanas, E. (2016). Price transmission in the international skim milk powder markets. Applied Economics, 48(54), 5233–5245.
Gervais, J.-P. (2011). Disentangling nonlinearities in the long- and short-run price relationships: An application to the US hog/pork supply chain. Applied Economics, 43,
1497–1510.
Goodwin, B. K., & Harper, D. C. (2000). Price transmission, threshold behavior, and asymmetric adjustment in the US pork sector. Journal of Agricultural & Applied
Economics, 32(3), 543–554.
Goodwin, B. K., & Holt, M. T. (1999). Price transmission and asymmetric adjustment in the us beef sector. American Journal of Agricultural Economics, 81(3), 630–637.
Granger, C. W. (1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica, 424–438. Journal of the Econometric Society.
Granger, C. W. (1988). Causality, cointegration, and control. Journal of Economic Dynamics and Control, 12(2–3), 551–559.
Greene, J. L. (2015). USDA’s ‘GIPSA Rule’on livestock and poultry marketing practices. In Congressional research service (Vol. 7, p. 5700). CRS Report for Congress.
Greenwood-Nimmo, M., & Shin, Y. (2013). Taxation and the asymmetric adjustment of selected retail energy prices in the UK. Economics Letters, 121(3), 411–416.
Hacker, S., & Hatemi-J, A. (2012). A bootstrap test for causality with endogenous lag length choice: Theory and application in finance. Journal of Economics Studies,
39(2), 144–160.
Hahn, W. F., et al. (2004). Beef and pork values and price spreads explained. USDA Economic Research Service.
Key, N., & McBride, W. D. (2007). The changing economics of US hog production (Vol. 52). USDA-ERS Economic Research Report.
Kwiatkowski, D., Phillips, P. C., Schmidt, P., Shin, Y., et al. (1992). Testing the null hypothesis of stationarity against the alternative of a unit root. Journal of
Econometrics, 54(1–3), 159–178.
Norwood, F. B., & Lusk, J. L. (2018). Agricultural marketing and price analysis. Waveland Press.
Panagiotou, D., & Stavrakoudis, A. (2017). Vertical price relationships between different cuts and quality grades in the us beef marketing channel: A wholesale-retail
analysis. The Journal of Economic Asymmetries, 16, 53–63.
Panagiotou, D., & Stavrakoudis, A. (2018). Free-on-board and uniform delivered pricing strategies in pure and mixed spatial duopolies: The strategic role of
cooperatives. The Journal of Economic Asymmetries, 18, Article e00109.
Pesaran, M. H., & Shin, Y. (1999). An autoregressive distributed-lag modelling approach to cointegration analysis. Econometric Society Monographs, 31, 371–413.
Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289–326.
Phillips, P. C., & Perron, P. (1988). Testing for a unit root in time series regression. Biometrika, 75(2), 335–346.
Qiu, F., & Goodwin, B. (2012). Asymmetric price transmission: A copula approach. In 2012 Annual Meeting, August 12-14, 2012, Seattle, Washington, number 124906.
Agricultural and Applied Economics Association.
Rezitis, A. N. (2018). Investigating price transmission in the Finnish dairy sector: An asymmetric NARDL approach. Empirical Economics, 1–40.
Rezitis, A. N., & Tsionas, M. (2019). Modeling asymmetric price transmission in the european food market. Economic Modelling, 76, 216–230.
Shin, Y., Yu, B., & Greenwood-Nimmo, M. (2014). Modelling asymmetric cointegration and dynamic multipliers in a nonlinear ARDL framework. Festschrift in honor of
peter schmidt. Springer.
United States Department of Agriculture – Economic Research Service. (2017). Retail prices for beef, pork, poultry cuts, eggs, and dairy products. http://www.ers.usda.
gov/data-products/meat-price-spreads.aspx. (Accessed December 2017).
U.S. Government Accountability Office (GAO. (2009). U.S. Government accountability office–concentration in agriculture, GAO-09- 746R, june 30. http://www.gao.
gov/new.items/d09746r.pdf.
Xia, T. (2009). Asymmetric price transmission, market power, and supply and demand curvature. Journal of Agricultural & Food Industrial Organization, 7(1).
Zivot, E., & Andrews, D. W. K. (2002). Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. Journal of Business & Economic Statistics,
20(1), 25–44.

13

You might also like