Pak. J. Agri. Sci., Vol.

44(2), 2007

TESTING THE LAW OF ONE PRICE: COTTON MARKET INTEGRATION IN

PAKISTAN’S PUNJAB
Khalid Mushtaq*, Faisal Abbas**and Abedullah***
*Department of Agricultural Economics, University of Agriculture, Faisalabad, Palostam
**Centre for Development Research (ZEF), University of Bonn, Germany
***Dept. of Environmental & Resource Economics, University of Agriculture, Faisalabad, Pakistan
Corresponding author’s E-mail: khalidmushtaq69@yahoo.Com

Regional market integration in many agricultural commodities has been extensively studied for the insight it
provides into the functioning of such markets; such studies provide valuable information about the dynamics of
market adjustments, and whether there exist market imperfections, which may justify government intervention.
This study empirically estimated the degree of integration in cotton markets of Pakistan’s Punjab using the law of
one price (LOP) framework and cointegration analysis. Cointegration Results show that all seed cotton markets
are highly integrated in the long run. Regression results show that g o vern men t in ter venti on in te rms o f
se ed c o tton proc uremen t neg ative l y in flue nced the d egr ee o f mark e t inte gra tio n . The high
degree of market integration observed in this case is consistent with view that Punjab’s seed cotton markets are
quite competitive and provide little justification for extensive and costly government intervention designed to
improve competitiveness to enhance market efficiency.
Keywords: Market integration, cointegration, cotton prices, Punjab.

INTRODUCTION used as a gauge of the success of market liberalization

and structural adjustment policies in developing
In a decentralized economic system resource countries. Market integration leads to market
allocation takes place through price signals transmitted liberalization or price stabilization because of detailed
by the markets. In developing economies, there are transmission of incentives across the marketing chain.
several impediments to the efficient functioning of Government of Pakistan tries to stabilize the prices of
markets, particularly agricultural commodity markets. cotton through price supports or producer subsidy. If
These includes not good enough transportation markets are well integrated the government will
infrastructure, difficulties in access to market stabilize the prices in one key market and rely on
information, government-imposed restriction on arbitrage to produce the similar outcome in other
movement of goods between regions, government markets. This reduces the cost of stabilization
monopoly over the marketing and distribution system, considerably.
and poor enforcement of anti-trust regulation that result Market integration is subjectively viewed as long run
in price fixing and Oligopolistic market structures. phenomenon. It is present whenever a stable price
Overall market performance may be evaluated in the relationship is established. This means that spatial
term of price relationships. Co integration test can be prices can temporarily deviate from each other in the
used to examine the stability of price relationship. short run and still be consistent with the idea of an
Although the larger markets that are better connected integrated market. The concept of spatial arbitrage is to
with the transport and communication network are visualize traders buying in low priced market,
expected to be well integrated; the same cannot be transferring the item to a high priced market, and
said about the smaller, more remote markets. Market reselling the purchased good in different localities tend
integration refers to co-movements of prices and more towards equality and move together with each other in
generally, to the smooth transmission of price signals integrated markets. Markets that are not integrated
and information across spatially separated markets. may convey inaccurate price information that might
Market integration provides important information on distort production decisions and contribute to
how markets work. Such information helps the inefficiencies in product markets.
government to decide the extent to which it should Cotton has greater significance in Pakistan’s economy.
promote market development. If, for example locations It accounts for 8.2 percent to value added in agriculture
A, B, and C are well integrated, then the government and about 2 percent to GDP. It provides raw materials
may think of withdrawing from, or at least reduce, its to local textile industry; the surplus lint cotton is
efforts to influence the price setting process in those exported. The share of cotton in value added by major
locations. Degree of market integration has often been crops is about 13 percent. Cotton farming is the

364
 Mushtaq, Abbas and Abedullah

principal source of raw material for the textile industry, (Alexander and Wyeth, 1994) extends the Ravallion
employing about 40 percent of the industrial labor. model and uses a single-equation error correction
Cotton and cotton products contribute about 10 percent model to test causality between prices. The
to agriculture GDP and 60 percent to the foreign cointegration approach (Palaskas and Harriss-White,
exchange earnings of the country. Cotton production in 1993, and Alexander and Wyeth, 1994) is based on the
Pakistan is concentrated in two provinces, Punjab and first step of the Engle and Granger (1987) two-step
Sindh, which jointly account for more than 99 percent method. Estimating:
of total production (GOP, 2003-04). The main interest P1t = β1 + β2P2t + ut t = 1,…,n (2)
of studying price integration among local markets is to a test of long-run spatial market integration is
be able to identify sets of markets that lead other equivalent to testing the stationarity of the residuals, ut.
markets in the price transmission process. Little has
been done in the way of empirically evaluating actual MATERIALS AND METHODS
cotton market performance in Pakistan. Tahir and Riaz
(1997) studied the integration of agricultural commodity The approach adopted here is based on the LOP in (2)
markets in southeastern Punjab, Pakistan, using but follows the Sims’ (1980) vector autoregressive
weekly prices of wheat, cotton and rice from 1993 to (VAR) methodology, unlike single-equation methods,
1995. The results suggested that there were no market the exogeneity of one price is not imposed ex ante;
integration for all cotton markets in the short-run long-run market integration is examined using
because of the periodicity of data used in the study and Johanson's (1988) cointegration procedure. This
the integration involved two different stages in the approach incorporates important features of previous
processing chain (cotton lint and seed cotton). The models. First, both prices are determined by their
cotton markets were also not integrated in the long-run current and past values. Second, the null hypothesis of
except Bahalwalnagar. Moreover, smaller markets no cointegration between two prices is a test of the
were more likely to be isolated as compared to large LOP which holds if the null is rejected. Given
markets. cointegration, the null of perfect market integration is
Besides locating factors at macro level affecting the tested where a price change in one market leads to an
integration of the markets this study aims critically equivalent price change in the other; imperfect market
estimating the extent of market integration in seed integration occurs if the relationship is not strictly
cotton markets of Punjab, Pakistan using the law of proportional.
one price (LOP) framework. The structure of the paper A price series is often trended and can be made
is as follows: Section 2 discusses the LOP, Section 3 stationary by first-differencing, that is it is integrated of
discusses the empirical approach, Section 4 discusses order one, or I(1). In general, the OLS regression in (2)
the data and results, while Section 5 concludes. is spurious since it is based on the assumption that
both series are stationary (Harris, 1995, p.14). The
The Law of One Price and Market Integration
exception is when (2) is cointegrated where the prices
Richardson (1978) notes that the LOP is a test of move together so that a stable relationship between
market integration in period t and involves the them is maintained. Any short-run disturbance away
regression: from this relationship induces changes in the prices so
that the relationship is maintained in the long run. In
∆P1t = β + β2∆P2t + ut t = 1,…,n (1) this sense, cointegration implies that a meaningful
where β1 and β2 are parameters, ∆ represents first long-run equilibrium exists (Granger, 1988). Since a
differences where for example ∆P1t=P1t-P1t-1, and ut is cointegrating relationship cannot exist between two
an error term with the usual properties. If the joint prices which are integrated of a different order, it is
hypothesis that β1=0 and β2=1 is not rejected, the two necessary to test for their order of integration. The
prices are statistically identical and the LOP holds. subsequent test for cointegration is a formal test of the
Equation (1) can be estimated using the original price long-run equilibrium relationship between pair-wise
series or the series in logarithms. The former implies prices.
an absolute price difference as the maintained We begin by testing for the presence of unit roots in
hypothesis while the latter implies a proportional price the individual time series of each model using the
difference. Ravallion (1986) extends (1) by assuming augmented Dickey-Fuller (ADF) test (Dickey and
that price adjustment between markets takes time, and Fuller, 1981), both with and without a deterministic
using an error correction model, a nested test for short- trend. The number of lags in the ADF-equation is
run market integration is shown to be equivalent to a chosen to ensure that serial correlation is absent using
test of the LOP. The Granger-causality approach the Breusch-Godfrey statistic (Greene, 2000, p.541). If

365
 Testing the law of one price in cotton markets

two prices are integrated of the same order, different seed cotton markets of Punjab i.e., Rahim Yar
Johansen's (1988) procedure can then be used to test Khan, Multan, Bahawalpur, Okara, D.G.Khan and
for the LOP between them. The procedure is based on Vehari. Table 1 reports the unit root results using ADF
maximum likelihood estimation of the vector error tests both with and without linear trend. Both models
correction model (VECM): indicate that null of unit root cannot be rejected for all

∆z t = δ + Γ1∆z t −1 + Γ2 ∆z t − 2 + Λ + Γp −1∆z t − p +1 + πz t − p + Ψx t + u t (3)

where zt is a vector of I(1) endogenous variables, price series as the absolute values of the ADF
∆zt=zt-zt-1, xt is vector of I(0) exogenous variables, and statistics are well below the 95 percent critical value of
π and Γiare (n×n) matrices of parameters with Γi=-(I-A1- the test statistics. We carried out an additional test
A2-…-Ai), (i=1,…,k-1), and π=I-π1-π2-…-πk. This called the Ф3 test. The null hypothesis in Ф3 test is that
specification provides information about the short-run the variable observed have unit root with no trend
and long-run adjustments to the changes in zt through against the alternative that the variables are trend
stationary. The values of F-statistics for all six variables
ˆ and
the estimates of Γ i π̂ respectively. The term are below the 95% critical value of the Ф3 test (7.24);
πz t − k provides information about the long-run therefore we reject the alternative and accept the null
equilibrium relationship between the variables in zt. hypothesis that means that all six price series have unit
Information about the number of cointegrating root and no trend.
relationships among the variables in zt is given by the Table 1. The unit root results
rank of the π-matrix: if π is of reduced rank, the model
Non-
is subject to a unit root; and if 0<r<n, where r is the Trended Ф3
Prices Trended
rank of π, π can be decomposed into two (n×r) matrices Model test
Model
α and β, such that π=αβ' where β'zt is stationary. Here,
Rahim Yar Khan -2.45 -1.63 3.27
α is the error correction term and measures the speed
of adjustment in ∆zt and β contains r distinct Multan -2.15 -1.40 2.52
cointegrating vectors, that is the cointegrating Bahawalpur -2.53 -2.19 3.53
relationships between the non-stationary variables. Okara -2.09 -1.40 2.41
Johansen (1988) uses the reduced rank regression D.G. Khan -3.11 -2.57 5.16
procedure to estimate the α- and β-matrices and the
Vehari -2.15 -1.57 2.43
trace test statistic is used to test the null hypothesis of
at most r cointegrating vectors against the alternative Critical Value -3.47 -2.90 7.24
that it is greater than r. (95% confidence level)

Regression Analysis The first step of the Johansen procedure is to select

the order of the VAR for each price relationship. We
Cotton market integration over time may be influenced use the LR-statistic, adjusted for small samples (Sims,
by various factors i.e. procurement, road length and 1980), to test the null hypothesis that the order of the
per capita income. A regression analysis of the VAR is k against the alternative that it is five where
following equation is undertaken to analyze the effects k=0,1,2,…,5 and for all cases, k=1. Johansen’s
of these factors in the cotton market. The functional cointegration results are presented in Table 2. The
form of the equation can be written as follows: trace test results suggested that these six price series
CS = β0 - β1 PROC + β2 ROAD + β3 (4) are strongly cointegrated and converge to long run
Where; CS = Co integration test statistic in markets; equilibrium in the sense that Punjab seed cotton
PROC = Procurement of seed cotton; ROAD = Length market system is stationary in four directions and non-
of roads in kilometers; and PCI = refers to Per capita stationary in two direction. In other words, four prices
income. can be expressed in terms of the other two prices
means that prices in six seed cotton markets are fully
RESULTS AND DISCUSSIONS cointegrated as law of one price (LOP) holds. It
suggests that even though the regional markets are
geographically dispersed, and therefore, spatially
Monthly wholesale prices (Rs./40 kg) of seed cotton
segmented, spatial pricing relationships reveal that the
from October-February (1992-93) to October-February
prices are linked together indicating that all the seed
(2003-04) (60 observations) in logarithmic form are cotton exchange locations are in the same economic
used for this study. The study is confined to six market.

366
 Mushtaq, Abbas and Abedullah

Table 3 reports the pair wise relationship between the relationship between (Rahim Yar Khan- Bahawalpur),
markets. The results of pair-wise cointegration (Rahim Yar Khan- D.G. Khan), (Multan- Bahawalpur),
relationships indicates that Rahim Yar Khan, Multan, (Bahawalpur-Okara) and (Bahawalpur-Vehari) show
Bahawalpur, Okara, Vehari and D.G. Khan seed cotton that these markets are not integrated with each other.
markets are integrated but not strongly, the pair-wise

Table 2. Co integration Results–Trace Statistics

EEquation Tested Null Alternative Statistics
Rahim Yar Khan, Multan, Bahawalpur, r=0 r≥1 179.37 (102.56)
Okara, D.G.Khan, Vehari r≤1 r≥2 126.87 (75.98)
r≤2 r≥3 82.89 (53.48)
r≤3 r≥4 43.35 (34.87)
r≤4 r≥5 11.34(20.18)
r≤5 r≥6 4.94 (9.16)
Note: Critical values (95% confidence level) in parentheses.
Table 3. Pair-wise Co integration Results–Trace Statistics
Equation Tested Null Alternative Statistics
Rahim Yar Khan-Multan r=0 r≥1 42.51 (20.18)
r≤1 r≥2 4.20 (7.53)
Rahim Yar Khan-Bahawalpur r=0 r≥1 15.02 (20.18)
r≤1 r≥2 5.29 (9.16)
Rahim Yar Khan-Okara r=0 r≥1 45.12 (20.18)
r≤1 r≥2 4.17 (9.16)
Rahim Yar Khan-D.G.Khan r=0 r≥1 19.68 (20.18)
r≤1 r≥2 4.28 (9.16)
Rahim Yar Khan-Vehari r=0 r≥1 40.19 (20.18)
r≤1 r≥2 4.56 (9.16)
Multan- Bahawalpur r=0 r≥1 14.87 (20.18)
r≤1 r≥2 5.05 (9.16)
Multan- Okara r=0 r≥1 39.09 (20.18)
r≤1 r≥2 4.15 (9.16)
Multan- D.G.Khan r=0 r≥1 22.18 (20.18)
r≤1 r≥2 4.05 (9.16)
Multan- Vehari r=0 r≥1 44.78 (20.18)
r≤1 r≥2 4.90 (9.16)
Bahawalpur- Okara r=0 r≥1 13.70 (20.18)
r≤1 r≥2 4.73 (9.16)
Bahawalpur- D.G.Khan r=0 r≥1 39.79 (20.18)
r≤1 r≥2 5.37 (9.16)
Bahawalpur- Vehari r=0 r≥1 14.16 (20.18)
r≤1 r≥2 4.89 (9.16)
Okara- D.G.Khan r=0 r≥1 21.73 (20.18)
r≤1 r≥2 3.91 (9.16)
Okara- Vehari r=0 r≥1 44.64 (20.18)
r≤1 r≥2 4.53 (9.16)
D.G.Khan-Vehari r=0 r≥1 24.17 (20.18)
r≤1 r≥2 4.34 (9.16)
Note: Critical values (95% confidence level) in parentheses.

367
 Testing the law of one price in cotton markets

Regression analysis of factors influencing seed cotton SUMMARY AND CONCLUSIONS

market integration is conducted using three years
moving average procedure. For this purpose, In the present study an effort was made to assess the
cointegration tests were performed annually for the degree of market integration in various seed cotton
markets under study for the period of 1992-93 to 2003- markets of Punjab province, Pakistan. Cointegration
04.Cointegration statistics (CS) results using Maximal techniques were applied to the monthly wholesale
Eigenvalue test statistics are reported in Table 4. prices data of seed cotton. The results indicated that
These statistics can be considered as the measure of these seed cotton markets are strongly cointegrated
market integration over time, in which the larger the and converge to long run equilibrium in the sense that
statistic, the stronger the degree of market integration. Punjab seed cotton market system is stationary in four
Table 4. Annual Cointegration test result using directions and non-stationary in two directions. In other
Maximal Eigenvalue test words, four prices can be expressed in terms of the
other two prices means that prices in six seed cotton
Years CS CS 3 Years moving avg. markets are fully cointegrated as law of one price
1991 52.50 - (LOP) holds. The high degree of market integration
1992 52.12 52.33 observed in this case is consistent with view that
1993 52.36 52.31 Punjab’s seed cotton markets are quite competitive
1994 52.45 52.56 and provide little justification for extensive and costly
government intervention designed to improve
1995 52.87 52.66
competitiveness to enhance market efficiency. The
1996 52.65 52.64 results also indicate that certain pair wise markets are
1997 52.39 52.63 not well integrated with each other. The integration
1998 52.85 52.47 among these markets can be improved through the
1999 52.16 52.59 promotion of price information and communication
2000 52.75 52.43 facilities.
2001 52.37 52.58
REFERENCES
2002 52.63 52.60
2003 52.81 - Alexander, C. and J. Wyeth. 1994. Cointegration and
CS: Cointegration Statistics Market Integration: An Application to the
Indonesian Rice Market, Journal of Development
The regression analysis results are shown in Table 5. Studies, 30, 303-328.
The results indicate that government intervention in Dickey, D.A. and M.A. Fuller. 1981. Likelihood Ration
terms of seed cotton procurement significantly Statistics for Autoregressive time series with a Unit
influenced the degree of market integration, implying Root, Econometric 49: pp (1057-1072).
that larger the seed cotton procurement, the lower the Engle, R.F. and C.W.J. Granger. 1987. Cointegration
degree of market integration. These results suggest and error correlation. Representation, Estimation
that procurement programs are less affective in and Testing, Econometrica 55: pp (251-276).
dynamic price adjustment. In addition, road length and Govt. of Pakistan. 2003-04. Economic Survey of
per capita income variables are not statistically Pakistan 2003-04. Finance Division, Economic
significant. Adviser’s Wing, Islamabad.
Table 5. Regression parameter estimates of factors Granger, C.W.J. 1988. Some Recent Developments in
influencing cotton market integration in a Concept of Causality, Journal of Econometrics,
Punjab, Pakistan 39, 199-211.
Variable Coefficients t-Value Greene, W.H. 2000. Econometric Analysis. New
Jersey: Prentice-Hall, Inc.
Intercept 52.182 92.667
Harris, R. 1995. Using Cointegration Analysis in
Procurement -0.145 -2.639* Econometric Modelling. London: Prentice Hall-
Road 0.000001396 0.425 Harvester Wheatsheaf.
Per capita income 0.000004377 0.424 Johansen, S. 1988. Statistical Analysis of cointegration
R2 0.642 of vectors, Journal of Economic Dynamic and
Control 12: pp 231-254.
Adjusted R2 0.508
*
Indicates significance at 95% confidence level

368
 Mushtaq, Abbas and Abedullah

Palaskas, T.B. and B. Harriss-White. 1993. The Richardson, J.D. 1978. Some Empirical Evidence of
Identification of Market Exogeneity and Market Commodity Arbitrage and the Law of One Price,
Dominance by Tests Instead of Assumption: An Journal of International Economics, 8, 341-352.
Application to Indian Material, Journal of Sims, C.A. 1980. Macroeconomics and Reality,
International Development, 8, 111-123. Econometrica, 48, 1-49.
Ravallion, M. 1986. Testing market integration, Tahir, Z. and K. Riaz. 1997. Integration of agricultural
American Journal of Agricultural Economics, 68(2): commodity markets in Punjab, the Pakistan
pp 292-307. Development Review. 36(3):241-262.

369