This paper aims to prove the impact of the recent oil roar on the public presentation of grain markets in Niger. In this state conveyance costs constitute the majority of trade costs, so the spike in fuel monetary value may negatively impact trade therefore taking to deficits and monetary value tensenesss on local grain markets. Assuming that markets are spatially arbitrated, the monetary value spread between two spatially separated markets is modelled as a nonlinear map of dealing costs. Compared to standard theoretical accounts, dealing costs and the velocity of accommodation to long-term equilibrium are allowed to change with the fuel monetary value. A threshold panel theoretical account is estimated on a sample of 66 market braces covering the period January 1990-October 2008. The consequences show that the accommodation velocity of monetary values reduces when the fuel monetary value additions. As a consequence, the hypothesis of a lag in the grain trade following the oil roar can non be rejected.

Cardinal words: Transaction costs, grain markets, oil roar, Niger, threshold panel theoretical account.

Introduction

On the international markets oil monetary values increased four times in current dollars over the period 2003-2008, conveying about an addition in fuel monetary values to Nigerien consumers of around 60 % for gasoline and 90 % for Diesel. This sudden rise had reverberations on all economic activity, in peculiar the conveyance industry, and as a consequence, on trade flows. In Niger, the grain trade involves long distances and the conveyance cost is the largest point in dealing costs. Aker ( 2007 ) estimates that the cost of fuel represents 85 % of conveyance costs. It is hence to be feared that the addition in oil monetary values will do a lag in trade as activity becomes less profitable, and will worsen the tensenesss on the grain markets.

If the markets are arbitrated, bargainers play a regulation function by transporting grain from excess parts where monetary values are low to shortage parts where monetary values are high, provided that the monetary value differences from one part to the other cover their dealing costs. At equilibrium, the arbitrage net income chances are void and the monetary values of a homogeneous merchandise such as millet are bound by a stable long-run relationship. In other words, at equilibrium, spacial arbitrage brings down the monetary value spread of transportation costs between two markets.

An addition in transportation costs – because of a rise in fuel monetary values for illustration – will non modify the relationship between market monetary values every bit long as trade remains profitable. The rise in dealing costs so generates an addition in the monetary value difference but the markets remain incorporate. However, the equilibrium relationship between monetary values on two different markets may be broken if the monetary value spread between these markets no longer covers the dealing costs. Minutess cease and the monetary values on each market vary independently of each other: the markets are segmented.

Owing to the deficiency of information about trade flows between the markets and about the different constituents of transportation costs other than fuel monetary values, it is non possible to straight prove the impact of the oil monetary value roar on the integrating of the millet markets. The attack therefore consists in proving the premise of a decrease in the velocity of accommodation of monetary value disequilibria to the rise in fuel monetary values.

In this attack the addition in oil monetary values is likened to a systemic daze which affects all the millet markets in the same manner. To prove this structural consequence of the oil monetary value rise, the millet monetary value accommodation theoretical account is estimated utilizing a threshold panel theoretical account. The sample is made up of 66 market braces and covers the period running from January 1990 to October 2008.

The consequences show that the velocity of accommodation of monetary value spreads to their long-term value diminishes during periods of high oil monetary values but ne’er beads to zero, while the dealing costs addition. In other words, monetary value disequilibria take clip to come down when the oil monetary value is high, but the monetary value spreads remain stationary. Although these consequences indicate a lag or even ephemeral or local discontinuities in trade during oil roar periods, they besides demonstrate the efficiency of the trade sector and make non govern out the hypothesis of grain market integrating over the survey period as a whole.

The first portion develops the millet market price-equilibrium relationship ensuing from the spacial arbitrage regulation. The 2nd portion presents the informations and the panel. Part Three presents the threshold panel theoretical account and Part Four the appraisal consequences. The concluding portion concludes.

1. Price equilibrium on grain markets

There is an copiousness of literature devoted to the spacial integrating of markets in developing states. Depending on the survey, the term “ market integrating ” may mention to extremely diverse state of affairss. Barrett and Li ( 2002 ) separate six possible market constellations runing from “ perfect integrating ” to “ metameric disequilibrium ” , depending on whether the markets are linked by trade or non and whether the arbitrage border is positive, negative or zero.

However, most of the literature, including this survey, is restricted by the absence of informations on trade and on dealing costs. Since Ravallion ( 1986 ) , most trials have been based on the parametric quantities of the equilibrium relationship between the monetary values of the same merchandise on geographically distinguishable markets when trade occurs.

This relationship is expressed in the signifier of a dynamic distributed slowdown theoretical account ( Ravallion, 1986 ) or, more frequently, as an error-correction theoretical account ( Ardeni, 1989 ; Alexander and Wyeth, 1994 ; Dercon, 1995 ; Palaskas and Harriss-White, 1993 ) . The dependent variable is the monetary value of the merchandise on a “ peripheral ” market[ 1 ]or the monetary value spread between two markets. The market efficiency or integrating trials are so based on steps of the transmittal of monetary value fluctuations from one market to the other, short-term and long-term, or the velocity of accommodation of the monetary value spreads to their long-term mark.

First expressed linearly, these theoretical accounts have been amended, notably under the influence of the plants by Spiller and Wood ( 1988 ) and Sexton, Kling and Carman ( 1991 ) , to take history of the trade discontinuities related to dealing costs. These writers, followed by Baulch ( 1997 ) among others, developed exchanging arrested development theoretical accounts to mensurate the grade of market integrating via the chances associated with each monetary value government ( Fafchamps and Gavian, 1996 ; Barrett and Li, 2002 ; Araujo et Al. 2005 ) . In parallel, threshold effects were introduced into the mistake rectification theoretical accounts, leting the velocity of accommodation of monetary value spreads to change harmonizing to whether the monetary value spread is inside or outside the dealing cost set ( Obstfeld and Taylor 1997, Abdulai, 2000 ; Von Cramon-Taubadel, 1998 ; Goodwin and Pigott, 2001 ) .

The analysis below comes in the continuity of these plants, taking to gauge market integrating through the velocity of accommodation of monetary value spreads between markets. Unlike in most old surveies, the dealing costs are non considered as fixed ; they can change harmonizing to the oil monetary value. In other words, the government exchanging procedure depends on the fuel monetary value degree.

The spacial arbitrage regulation

Harmonizing to the spacial arbitrage regulation, the monetary value spread of the same merchandise on two geographically distinguishable markets, I and J, is at most equal to the dealing costs involved in transporting the merchandise from one market to the other, i.e. :

A?Pi – Pj A?i‚? rij ( 1 )

Pi and Pj are the millet monetary values on markets one and J

rij: dealing costs between I and J

The dealing costs between I and J include non merely conveyance costs but besides hunt and information costs, bargaining and contract enforcement costs, bargainers ‘ border, hazard premium, etc. By their very nature, a big proportion of these costs are non mensurable.

Equation ( 1 ) is known as the look of the “ weak signifier ” of the jurisprudence of one monetary value ( LOP ) . If the markets are arbitrated, the monetary value spread between two markets may be lower than the dealing costs, but it can non be higher. If the monetary value spread between I and J is purely lower than the dealing costs, so trade is non profitable and no trade occurs between I and J: the markets are segmented.

When trade occurs, the monetary value spread between markets one and J is equal to the dealing costs, i.e. :

A?Pi – Pj A?= rij ( 2 )

Equation ( 2 ) gives the “ strong signifier ” of the jurisprudence of one monetary value. The markets are bound by commercial minutess and any monetary value fluctuation on one market is transferred in full to the other market: the markets are said to be integrated.

Two extra hypotheses are by and large made in the literature on spacial market integrating. On the one manus, most writers assume the being of time-lags in monetary value accommodations due to the hapless quality of conveyance and communicating substructures in developing states. On the other manus, the dealing costs, which are an unobservable variable, are by and large assumed to be changeless.[ 2 ]

Equation ( 2 ) is therefore rewritten as a partial accommodation theoretical account in which the dealing costs are changeless and the long-term monetary value spread ( z* ) is given by:

( 3 )

and: and 0 i‚? i?¬ i‚?1 ( 4 )

with:

: equilibrium monetary value spread

degree Celsiuss: dealing costs

i?¬ : velocity of accommodation

i?? ~ iid ( 0, i??A?i?? )

From ( 3 ) and ( 4 ) comes:

( 5 )

Equation ( 5 ) gives the equilibrium relationship between monetary values when the markets are integrated. To take the nonlinear nature of the spacial arbitrage regulation into history, threshold effects can be introduced into this equation. Most surveies so make the accommodation velocity of disequilibria ( i?¬ ) depend on the price-spread degree in the old period ( zt-1 )[ 3 ]( see for illustration Obstfeld and Taylor, 1997, Lo and Zivot, 2001 ) .

If the monetary value spread is big, higher than threshold degree Celsius, trade is profitable, and the monetary value disequilibria are adjusted:

if zt-1 a‰? degree Celsius ( 6a )

However, if the monetary value spread in absolute value is little, lower than threshold degree Celsius, arbitrage is non profitable, i?¬ = 0, and the monetary value spread follows a random walk:

if zt-1 & lt ; degree Celsiuss ( 6b )

In this attack the dealing costs are changeless and the monetary value spread may, at least transitorily, be higher than the dealing costs.

Impact of the rise in fuel monetary values

The graduated table of the rise in oil monetary values in recent times and the preponderance of conveyance costs in dealing costs in Niger lead us to reconsider the hypothesis that dealing costs are fixed and to look at the impact of this rise on market integrating. Whilst keeping the premise that markets are spatially arbitrated, the impact of the rise in oil monetary values on the integrating of grain markets in Niger may be estimated by agencies of an equation derived from ( 5 ) .

Compared with the old attack, any monetary value spread value is compatible with the market integrating hypothesis provided that the dealing costs are variable and non-observable. In other words, the monetary value spread does non give information about whether the markets are integrated or non. However, for unchanged local supply and demand conditions on markets one and J, an addition in the dealing costs diminishes the likeliness of market integrating. More specifically, two different effects of the rise in oil monetary values on market integrating are expected harmonizing to the graduated table of this rise.

If the oil monetary value ( qt ) and accordingly the dealing costs remain moderate ( with an oil monetary value lower than threshold i?§ ) , trade is still profitable. In this instance bargainers are expected to go through on the addition in their dealing costs to the sale monetary value of the merchandise. The spacial scattering of monetary values ( measured by the monetary value spread between two markets ) increases but the velocity of monetary value accommodation remains unchanged. We get:

, 0 & lt ; i?¬ i‚? 1, if qt i‚? i?§ ( 7a )

with

Nutmeg State: the dealing costs that can be approached by a vector of variables including, among other things, the oil monetary value ( qt ) .

i?µ ~ iid ( 0, i??A?i?µ )

However, once more for unchanged market conditions, if the rise in dealing costs is big, such that bargainers can non go through it on to consumer monetary values, trade, which has become less profitable, is suspended:

, i?¬ = 0, if qt & gt ; i?§ ( 7b )

To prove the hypothesis whereby the rise in fuel monetary values may be the ground for trade discontinuities and to place the critical oil monetary value degree above which the markets no longer adjust, we use a threshold panel theoretical account. The information and the panel construction are foremost presented before we develop the econometric theoretical account and the consequences.

2. Datas

The millet monetary value informations used semen from the Agricultural Market Information System ( SIMA ) of Niger ; 12 market monetary values are included covering the period from January 1990 to October 2008.

Table 1. Niger, millet monetary values. Observation period: 1990.01 – 2008.10

Markets

Mean

( Fcfa/kg )

Minute

( Fcfa/kg )

Soap

( Fcfa/kg )

Coefficient of fluctuation

Agadez

134

52

337

0.40

Bakin Birgi

96

32

292

0.46

Birni Konni

118

50

281

0.40

Dosso

139

58

329

0.40

Gaya

124

42

315

0.43

Katako

141

71

324

0.35

Koundoumawa

98

32

275

0.48

Maradi

104

39

261

0.44

Tahoua

144

54

369

0.40

Tillaberi

145

58

306

0.37

Tounfafi

114

42

273

0.44

Zinder

109

40

312

0.43

Beginning: Agricultural Market Information System ( SIMA ) , Niger.

Katako is the chief grain market of Niamey.

These markets have been chosen from among all the markets monitored by the SIMA ; they were selected harmonizing to the figure of losing informations ( less than 5 over the period as a whole ) ,[ 4 ]the size of the markets in footings of supply and demand, and whether they are located in a excess or shortage part. Of these 12 markets, Niamey ( Katako ) and to a lesser extent Agadez are large-consumption markets that exert regional influence ( Araujo et al. 2008 ) . Birni Konni, Maradi, Koundoumawa and Zinder are located in the chief millet production zones near to the boundary line with Nigeria, Niger ‘s taking beginning of grain imports. Bakin Birji is a market near to Zinder, located in a production country that is considered as medium-vulnerable. Gaya and Tillaberi are boundary line markets, the gateways for cereals imported from Benin and Burkina Faso severally. Gaya is the chief market in a part demoing a excess, while Tillaberi is in a shortage part that is highly vulnerable in times of crisis. Tahoua is the capital of a part that includes Tounfafi, structurally in shortage and besides considered as extremely vulnerable to provide and monetary value dazes ; Dosso is the capital of a vulnerable part in shortage.[ 5 ]

Graph 1. Real monetary values of millet on eight Nigerien markets ( Fcfa/kg )

January 1990-October 2008 ( establish 100=2000 )

Graph 2. Price spreads in existent footings for three market braces ( Fcfa/kg )

January 1990-October 2008 ( establish 100=2000 )

Millet is the chief grain consumed in Niger. Around 70 % of ingestion is covered by national production, with the remainder imported from Nigeria, Burkina Faso and Mali. Millet production chiefly comes from little, non-mechanized farms that use small if any fertiliser or phytosanitary merchandises. As a consequence, labor costs account for the bulk of millet production costs.

Millet monetary value fluctuations ( chart 1 ) reflect the clime dazes and parasite onslaughts that on a regular basis hit Nigerien agribusiness and coincide with episodes of acute malnutrition, or even famine. The biggest production crises were recorded in 1997, 2000 and 2001. The 2005 monetary value rise was foremost and foremost the consequence of a supply deficit in Niger and the states in the sub-region, peculiarly Nigeria ( Michiels and Egg, 2007 ) .

The inter-market monetary value spreads fluctuate greatly as illustrated by Graph 2 but seem to demo a decrease in dealing costs at the start of the period under survey ( 1990-96 ) , followed by cyclical fluctuations.

Graph 3. Oil and fuel monetary values in Niger ( existent CFA francs )

Since 1998, the growing of fuel monetary values in Niger has rather closely followed that of oil monetary values on the international market ( Graph 3 ) . However, in the period 1993-97, there was a major divergency between fuel monetary values on the Niger market and the oil monetary value expressed in changeless CFA francs. Petrol and Diesel monetary values fell in current value in 1993 and stayed invariable until 1997, despite the devaluation of the CFA franc in January 1994. Between 1994 and 1997 Nigerien consumers therefore benefited from a fuel monetary value subsidy of the order of 40 % .[ 6 ]The low monetary values on the Nigerien market in that period correspond to an alliance of monetary values with those in Nigeria ( Herrera, 1998 ) .

All in all, three stages can quite clearly be distinguished in the growing of fuel monetary values in Niger: a period of really low monetary values ( 1994-97 ) , followed by a catch-up stage ( 1998-2003 ) , so an addition caused by the rise in monetary values on the international market ( 2004-2008 ) . This led us to prove for the being of at least three governments in the equilibrium relationship between millet monetary values on geographically distinguishable markets.

In the econometric analysis that follows, the monetary value of gasoline in Niger is considered as representative of the cost of fuel in this state. However, this variable is merely a placeholder variable for the existent monetary value of fuel in Niger. In Niger there is a parallel market for crude oil merchandises from Nigeria and this market is peculiarly active in boundary line zones. Unfortunately, the deficiency of information prevents us from ciphering a fuel monetary value index that takes the being of this parallel market into history. The premise is made that the growing in fuel monetary values in Niger follows that of fuel in Nigeria, and/or that the bulk of grain bargainers do non fall back to the informal fuel market.

Table 2. Panel unit root trial: period 1990.02 – 2008.11 ; 66 market braces

Trials

Millet monetary value

Monetary value spread

Petrol monetary value

UR common to all markets

Levin – Lin – Chu

– 2.39

( 0.01 )

-82.38

( 0.00 )

Breintung

– 1.58

( 0.05 )

-43.36

( 0.00 )

UR specific to each market

Im – Pesaran – Shin

-5.23

( 0.00 )

-65.86

( 0.00 )

15.46

( 1.00 )

Maddala – Wu ( ADF )

323.3

( 0.00 )

-3399.8

( 0.00 )

4.55

( 1.00 )

Maddala – Wu ( PP )

280.01

( 0.00 )

-4073

( 0.00 )

3.18

( 1.00 )

Choi ( ADF )

– 4.23

( 0.00 )

-53.73

( 0.00 )

14.83

( 1.00 )

Choi ( PP )

– 3.66

( 0.00 )

-60.15

( 0.00 )

16.09

( 1.00 )

LLC and Breintung: Holmium: common unit root for all the markets. The Breintung trial is conducted on series that are free of autocorrelations and deterministic constituents. .

IPS, MW and Choi: Holmium: unit root specific to each market brace. Information science: statistic based on ADF. MW and Choi: statistic based on ADF or PP.

P-values in parentheses.

The informations are organized into a panel composed of 66 market braces and 225 monthly periods. The unit root trials conducted on the panel informations reject the presence of a unit root in the millet monetary value series and take us to see this series as stationary ( Table 2 ) . The same goes for monetary value spread znt. However, the presence of a unit root in the fuel monetary value series can non be rejected.

This consequence may intend that the millet monetary value spreads vary independently of the monetary value of gasoline and that the markets are segmented. It may besides intend that there are other, non-stationary constituents of dealing costs, cointegrated with the monetary value of gasoline and non-observed. In this instance – and this is the preferable hypothesis here – the non-stationarity of gasoline monetary values remains compatible with the market integrating hypothesis.

3. Econometric theoretical account

To prove for the being of different monetary value equilibrium regimes harmonizing to the fuel monetary value degree, we use a threshold panel theoretical account following the attack by Hansen ( 1999 ) . This patterning serves to nail interruptions in the estimated relationship and to screen the observations into different categories harmonizing to the value taken by the threshold variable, here the fuel monetary value on the Niger market.

In the absence of informations on dealing cost constituents other than the fuel monetary value, it is non possible to gauge straight the nexus between the monetary value spread of millet, a stationary variable, and the monetary value of gasoline, an integrated variable of order one. This leads us to do the premise that the dealing costs are changeless per interval, and adjust bit by bit to the fluctuations in oil monetary values. In this hypothesis, merely major fuel monetary value fluctuations are passed on to the dealing costs.

The general signifier of the estimated theoretical account, with m-1 thresholds ( i?§m-1 ) finding thousand monetary value governments, is given by:

Or, equivalently:

( 8 )

with:

znt: the monetary value spread in absolute value for single n. T = 1, aˆ¦ , T. T = 225

n represents a market brace, n = 1, aˆ¦ , N. N= 66.

i?­sA = i?¬scsA and Cs: average dealing cost in government s ; s = 1, aˆ¦ , m.

i?¬s: accommodation velocity in government s.

i?§ ‘ = ( i?§1, aˆ¦ , i?§m-1 ) and ( i?§1 & lt ; i?§2 & lt ; aˆ¦ & lt ; i?§m-1 ) the vector of the m-1 threshold parametric quantities finding thousand monetary value governments.

Is ( i?§ ) the Heaviside map, such that Is ( i?§ ) = I ( i?§s-1 & lt ; qt i‚? i?§s ) = 1 if i?§s-1 & lt ; qt i‚? i?§s et Is ( i?§ ) = 0 otherwise. s = 0, aˆ¦ , m, i?§0 = 0 and i?§m = + i‚?

qt is the threshold variable mensurating the unit conveyance cost by the monetary value of gasoline in Niger.

i?¤n is the fixed portion of dealing costs, specific to each market brace.

i?±t is the fixed portion of dealing costs, specific to each period.

ent ~ iid ( 0, ) .

In this theoretical account, the dealing costs ( degree Celsius ) and the accommodation velocity of monetary value spreads ( i?¬ ) vary harmonizing to the oil monetary value ( qt ) . In the government matching to high conveyance costs, the velocity of accommodation i?¬ is expected to be given towards zero.

In equation ( 8 ) , the lagged variable znt-1 is positively correlated with the single effects ( i?¤n ) ( specific to each market brace ) . However, in the present instance the clip deepness of the sample should be big plenty ( 225 monthly observations ) for the intra-individual calculator of i?¬ to be convergent ( Nickell, 1981 ; Kiviet, 1995 ) .

The coefficients of equation ( 8 ) and the threshold parametric quantities i?§ are therefore estimated utilizing the within calculator. The within calculator of i?§ is given by ( Hansen, 1999 ) :

with, the amount of squared remainders.

Threshold value i?§ is sought from among the values of qt such that each government contains at least 5 % of the observations. Given the high figure of possible values of ( T*0.90 ) , the hunts are restricted to integer values of qt.

The thresholds are identified by following the consecutive process presented in Hansen ( 1999 ) . Once the first threshold has been identified, the estimation of the 2nd threshold is given by:

with:

After gauging the vector, the coefficients estimations are given by and

To prove the hypothesis of a nonlinear theoretical account with g governments against that of a theoretical account with thousand governments ( K & gt ; g ) , the Fgk statistic is calculated:

A

As the asymptotic distribution of Fgk is non standard, the critical values must be bootstrapped. We follow the Hansen process, which consists of sing regressors ( znt-1 ) and threshold variable ( qt ) as given, and maintaining them fixed in the bootstrap samples.

Hansen ( 1999 ) showed that is a convergent calculator of true threshold value i?§s0 and that its asymptotic distribution is non standard. To prove H0: i?§s = i?§s0, the likeliness ratio trial is used, with critical values c ( i?? ) of LRs ( i?§s0 ) given by Hansen ( 1999 ) and:[ 7 ]

The assurance intervals shown in the consequences table are given for i?? = 1 % and i?? = 5 % ; they include all the values of i?§ such that LRs ( i?§s0 ) A i‚? degree Celsius ( i?? ) .

4. Appraisal consequences

In the estimated theoretical account all the monetary values are expressed in existent footings, that is, deflated by the consumer monetary value index in Niger. Monthly silent person variables are introduced in order to command for seasonal effects. Annual silent person variables are besides introduced to command for one-year clip effects. A one-threshold theoretical account is foremost estimated ; if the threshold is important, a two-threshold theoretical account specifying three monetary value governments is estimated.

Table 3. Appraisal consequences for

Dependent variable: i?„znt

H0

( 1 )

( 2 )

( 3 )

( 4 )

( 5 )

i?­

0.134***

0.085***

0.114***

0.157***

0.083***

( 0.006 )

( 0.010 )

( 0.007 )

( 0.012 )

( 0.010 )

znt-1

-0.535***

-0.536***

( 0.011 )

( 0.011 )

I ( qti‚?i?§1 )

-0.086***

-0.061***

( 0.013 )

( 0.013 )

I ( qt & gt ; i?§2 )

0.045***

0.037***

( 0.007 )

( 0.008 )

znt-1I ( qti‚?i?§1 )

-0.625***

-0.622***

-0.603***

( 0.017 )

( 0.017 )

( 0.017 )

znt-1I ( i?§1 & lt ; qti‚?i?§2 )

-0.529***

-0.526***

-0.527***

( 0.012 )

( 0.012 )

( 0.012 )

znt-1I ( qt & gt ; i?§2 )

-0.470***

-0.465***

-0.499***

( 0.017 )

( 0.020 )

( 0.018 )

Rainfallt

-1.40E-04***

( 2.04E-05 )

Monthly silent persons

yes

yes

yes

yes

yes

Annual silent persons

yes

yes

yes

yes

yes

Individual fixed effects

yes

yes

yes

yes

yes

RA?adl

0.275

0.280

0.279

0.283

0.282

330

330

330

330

Assurance interval at 95 %

[ 330 ; 330 ]

[ 330 ; 332 ]

[ 330 ; 332 ]

[ 330 ; 330 ]

Assurance interval at 99 %

[ 330 ; 330 ]

[ 330 ; 332 ]

[ 330 ; 332 ]

[ 330 ; 330 ]

462

462

462

462

Assurance interval at 95 %

[ 462 ; 462 ]

[ 449 ; 462 ]

[ 449 ; 464 ]

[ 462 ; 462 ]

Assurance interval at 99 %

[ 449 ; 462 ]

[ 449 ; 464 ]

[ 449 ; 468 ]

[ 449 ; 462 ]

Nr Ob in the lower government

3168

3168

3168

3168

Nr Ob in the intermediate government

8712

8712

8646

8712

Nr Ob in the upper government

2970

2970

2112

2970

F12

56.60

41.18

34.27

94.4

( P-value ) a

( 0.6 )

( 0.02 )

( 0.03 )

( 0.00 )

F13

104.60

94.4

86.47

145.31

( P-value ) a

( 0.7 )

( 0.00 )

( 0.00 )

( 0.02 )

F23

47.90

53.08

51.46

50.63

( P value ) a

( 0.56 )

( 0.00 )

( 0.00 )

( 0.25 )

Wald Test i?¬2 = i?¬1

32.13

31.17

19.42

( P-value )

( 0.00 )

( 0.00 )

( 0.00 )

Wald Test i?¬1= i?¬3

53.60

43.08

24.16

( P-value )

( 0.00 )

( 0.00 )

( 0.00 )

Wald Test i?¬2= i?¬3

12.07

9.62

2.54

( P-value )

( 0.00 )

( 0.00 )

( 0.11 )

Nr observationsA

14850

14850

14850

13926

14850

In parentheses, the robust criterion divergences. *** : important at 1 %

a Bootstrap P-value on 500 reproductions

The appraisal consequences are presented in Table 3. The additive theoretical account matching to the void hypothesis of oil monetary values holding zero impact on the arbitrage relationship parametric quantities is given in column ( 1 ) . Alternate theoretical accounts are tested: in specification ( 2 ) the dealing costs vary harmonizing to the oil monetary value, but the accommodation velocity remains fixed. In specifications ( 3 ) , ( 4 ) and ( 5 ) the accommodation velocity and the dealing costs vary with the fuel monetary value. Specification ( 5 ) , which allows the changeless to be different in each monetary value government, is less restrictive than specifications ( 3 ) and ( 4 ) .

Irrespective of the specification tested, the two thresholds identified are the same. The first threshold settees at Fcfa 330 per liter of gasoline and the 2nd threshold at Fcfa 462 per liter ( 2000 existent monetary values ) . This 2nd threshold corresponds to the start of the recent oil monetary value roar. However, it is observed that the 2nd threshold, whose assurance interval is broader, is identified less exactly than the first ( Graphs 4 and 5 ) .

Graph 4. LR1 statistic for Graph 5. LR2 statistic for

the 1st threshold in equation ( 3 ) the 2nd threshold in equation ( 3 )

The parametric quantities estimated in the assorted specifications are in line with outlooks: the monetary value accommodation velocity diminishes and the dealing costs addition when the monetary value of gasoline exceeds the first and so the 2nd threshold. For illustration, the consequences in column ( 3 ) show that the velocity of accommodation of monetary value disequilibria drops from 0.62 to 0.53 so to 0.47 when the monetary value of gasoline goes past the first so the 2nd threshold, bespeaking an addition in average dealing costs from 0.18 to 0.21 so to 0.23 Fcfa/kg ( Table 4 ) . In column ( 5 ) , the accommodation velocity falls from 0.60 to 0.53 so to 0.50 and the average dealing costs, which were close to zero in the first government, rise to 0.16 so to 0.24 in the 2nd and 3rd governments ( Table 4 ) .

The increased continuity of dazes in the 2nd and 3rd governments leads to a lag in trade, or even a local and/or ephemeral cleavage of markets. However, the markets remain globally integrated by trade: the accommodation velocity does non be given towards zero and the rise in fuel monetary values is at least partly passed on to the monetary value spreads.

Harmonizing to the nonlinearity trials ( F-tests ) specifications ( 3 ) and ( 4 ) are preferred. Model ( 4 ) incorporates an excess variable compared to specification ( 3 ) . It is a clime variable measured by the average cumulative rainfall in the production zones during the cropping season. This variable serves to look into for the being of a common monetary value motion caused by the conditions conditions. Its negative coefficient shows that monetary values and the spacial scattering of monetary values addition when rainfall is low. Unfortunately, rainfall informations are non available for the terminal of the period under survey ( the rainfall series ends in September 2007 ) , such that the monetary value government estimates matching to a high oil monetary value are less precise in this specification.

Table 4. accommodation velocities and monetary value spreads in equation ( 3 )

Threshold ( Fcfa/litre )

330 462

Equation ( 3 )

i?¬

0.62

0.53

0.47

degree Celsiuss

0.18

0.21

0.23

Equation ( 5 )

i?¬

0.60

0.53

0.50

degree Celsiuss

0.04

0.16

0.24

0.25

0.27

0.27

degree Celsiuss = i?­/i?¬

All in all, the two thresholds identified determine three monetary value governments which schematically correspond to three periods ( Graph 6, Table 4 ) :

1994-1997: the monetary value of fuel in Niger is subsidized and maintained at a low degree ; the millet markets are good incorporate and the exogenic dazes are quickly absorbed.

1990-1993 and 1998-2005: the monetary value of gasoline is all of a sudden adjusted upwards in 1998 and in existent footings returns to a degree near to that at the start of the period ; spacial arbitrage is less efficient, the dazes last longer, and the average dealing cost additions.

2005 – 2008: the monetary value of gasoline reaches record degrees following the oil monetary value roar ; the velocity of accommodation beads further while the dealing costs addition ; the grain trade is even less reactive to monetary value disequilibria.

Graph 6. Gasoline monetary values and threshold values ( changeless CFA Francs )

An alternate modeling of the monetary value spreads, matching to dealing costs that are purely relative to the millet monetary value, was besides tested. In this instance the dependant variable in ( 8 ) is the comparative monetary value spread, in other words the border ratio, i.e. : . The consequences, which are presented in the appendix, do non govern out the hypothesis of a lag in the monetary value accommodation velocity during the oil monetary value roar. This specification of dealing costs however seems really restrictive.

5. Decision

Since 1994, fuel monetary values in Niger have risen aggressively, foremost in the aftermath of an accommodation of domestic monetary values to universe monetary values in 1998 and so the oil monetary value roar at the terminal of the period. In existent footings the monetary value of gasoline about tripled between 1997 and 2008. Paradoxically, although fuel is a major point in grain dealing costs in Niger, the correlativity between the monetary value spread between market braces, a stationary series, and the fuel monetary value, an incorporate series of order one, is negligible.

However, although there is no direct nexus between fuel monetary value fluctuations and grain selling costs, the analysis conducted here can non govern out the hypothesis of a gradual accommodation of the dealing costs to fluctuations in fuel monetary values. In other words, bargainers merely pass on the large fluctuations in oil monetary values to millet consumers, and adjust the other constituents of their dealing costs in response to low fluctuations in fuel monetary values.

But the analysis besides shows a autumn in the velocity of accommodation of millet monetary values between the different markets during periods of high oil monetary values. The greater continuity of monetary value disequilibria over the last period shows a lower responsiveness of trade to monetary value disequilibria. We therefore can non reject the hypothesis of a planetary lag in trade and a reorientation of trade flows towards the most profitable markets during the period matching to the last oil monetary value roar. Merely a more disaggregated market-by-market analysis could place the markets that are transitorily isolated during the oil monetary value roar.

The analysis reaches the decision of a good planetary public presentation of the grain markets, which remain incorporate despite the important addition in fuel monetary values at the terminal of the period. Like most analyses carried out in developing states, the range of the consequences is limited by a deficiency of informations about trade and the assorted constituents in dealing costs. It is besides possible that the different monetary value governments are falsely identified if the threshold variable used – the official monetary value of gasoline – does non decently reflect the growing in the monetary value of fuel used for the grain trade in Niger. More by and large, these consequences point to a demand for farther research into the nature of dealing costs and trade schemes in Niger.