Etude de stabilitÃ© de la relation entre Inflation et ...

Empirical Economics Review 2(1): (December 2011) ISSN 2222-9736

Is there a Stable Relationship between Inflation and Capacity Utilization in Tunisia? Kamel Helali Department of Applied Quantitative Methods, Faculty of Economics and Management of Sfax, Tunisia Email: [email protected] Abstract: In recent years, the decline in the monetary aggregates, as an effective indicator for making a future forecast on inflation, has led the researchers in finance and economists to control the economic statistics. To better handle the exchange between inflation and the economic activity, these researchers have developed a distinction between the unemployment rate and the Non-Accelerating Inflation Rate of Unemployment (NAIRU), where the former does not generate an increase of inflation. Similarly, we can use the capacity utilization rate (CU), instead of the unemployment rate, as an inflationary pressure indicator to estimate a balance rate or the Non Accelerating Inflation Capacity Utilization (NAICU). Typically, from this rate of balance, we notice an increase of inflation and a beginning of production obstruction. In this survey, we try to find out if there is a similarity between the analyses using the concept of NAIRU and that of NAICU. The appropriate means for studying this similarity is the Phillips curve of short run where we substitute the unemployment rate by the CU. The increased Phillips curve states a negative exchange between inflation and unemployment for a given level of anticipated inflation. It is necessary to note that the natural unemployment rate is also considered as a long term unemployment rate which the economy tends to follow through time. Keywords: Capacity utilization, Inflation, NAICU, NAIRU, ARDL JEL Classification Number: E22, E23, E24, E31

1. Introduction Several economists set different questions about the use of the CU as a measure of productive resources. The first question turns around the definition of the used CU. Generally, the theoretical definition developed by Berndt and Morrison (1981) is the most used in the different surveys. The second question is about the strong relationship between inflation and the CU. It is necessary to mention that an elevated CU can decide the policy of an economy in relation to its inflation rate. The third question is based on the fact that the CU and the unemployment rate are complementary or substitutable in the inflation equation. The use of the CU as a sign of inflation appearance is an advantage for the economic activity. The inflationary pressures appear typically when there is a substantial increase of

Empirical Economics Review 2(1): (December 2011)

38

aggregated demand of goods and services compared to the aggregated offer, thus causing either a decrease of quantities of non used productive resources. Economists measure this lack in different ways. The most used measures are the unemployment rate, the gap of unemployment rate, the real production gap, and the CU (Rennison, 2003). In fact, are the CU and the unemployment complementary or substitutable in the inflation equation? In addition, if the CU is an efficient instrument of inflationary pressures, is there a level of CU above which the politicians become interested in the potential of inflation acceleration? According to several authors, namely McElhattan (1978 and 1985), Shapiro (1989), Gordon (1989), Garner (1994), Finn (1995 and 1996), Corrado and Mattey (1997), Belton and Cebula (2000) and Nahuis (2003), the relation between the CU and inflation is very important. The CU seems to be an efficient indicator of inflationary pressures. Although the relative ability of unemployment rate and the CU in the explanation of inflation has been largely discussed by economists, few formal empirical researches have been produced to lighten them. Particularly Finn (1996) represents the unique direct empirical study to invest in this question. Thus, she breaks up the CU in two components, one is high and the other is low in relation to the rate of balance. She shows, indeed, that only high rates can have a meaningful effect on inflationary pressures. 2. Theoretical Development of the Relationship Inflation-Capacity Utilization The underlying idea of this theory is that there is a relationship between the CU and inflation deducted from simple economic notions. When there is a non-used capacity, competition of producers’ leads to a decrease of prices. The appearance of capacity constraints is going to cause an increase of pressures on competitions and prices. Consequently, an economy facing these constraints has to expect an increase of the supply prices following the demand excess. The situation described here is analogous to Phillips curve, which postulates a relation between an anticipated real salary and unemployment. When several agents are unemployed, salaries and prices do not increase well mainly when firms dispute in order to solve the unemployment problem. 2.1. The NAIRU Concept The NAIRU is a very important economic and political concept. It fixes for each Nation, at a given time, the minimum unemployment rate which can be accompanied by a stability of prices. These recommendations can show governments what minimum unemployment rate they have to maintain in order to prevent an increase of salaries that would cause inflation. The concept is related, but not equivalent, to that of the natural unemployment of Milton Friedman.


39

By referring to Ball and Mankiw (2002) and Sekhon (2001), we are going to develop a simple model permitting to estimate the NAIRU of the Tunisian economy. It is important to note that it is only significant since it plays an essential role to guide the economic policy. The NAIRU model, as a standard model of inflation based on the anticipation of augmented Phillips curve, is:

(

)

π t − π te = β u t − u*t + δX t + ζ t where π t is the current inflation rate, π te is the anticipated inflation rate, u t is the

ut* is the NAIRU, X t contains additional explanatory variables to

unemployment rate,

control offer shocks, and ζ t is a random error. In reference to the works of Gordon (1995), Tootell (1994), and Weiner (1993), we use a random step process to give an anticipation of inflation. Thus π te = π t −1 , therefore π t − π te = ∆π t . Beyond, equation (1) will be written:

(

)

∆π t = β ut − u*t + γX t + ζ t This equation neglects the possibility of serial correlation in terms of errors. Consequently, it is conventional to estimate an autoregressive specification:

(

)

∆π t = β (L ) u t − u*t + δ (L )∆π t −1 + γ (L )X t + ε t where « L » is an operator of delay, β (L ) , δ (L ) , and γ (L ) are polynomial lags and ε t is a non correlated error term. Equation (3) represents the difficulty in estimating this model since it is non-linear in parameters. When the NAIRU,

ut* , no longer varies over time. This equation can be

written in a form that can be estimated conventionally by the Ordinary Least Squared:

∆π t = µ + β (L )ut + δ (L )∆π t −1 + γ (L )X t + ε t The NAIRU estimator is then:

uˆ* =

where

− µˆ ˆ β (1)

β (1) = ∑ip=1 β i , with p number of lag of the polynomial β (L ) . Note that the

NAIRU is a non linear function of coefficients µ and β (1) .

40


The approach estimated in equations (4) and (5) can change in case where the NAIRU varies with time. In order to accomplish this work, we have to replace the µˆ by

∑ j =1 α j H j −1 (t ) , l

where H j is a Hermite polynomial of order j and « t » is the time

centred around 0. This is explained by the fact that the Hermite polynomials are orthogonal and symmetrical compared to the interval (− ∞;+∞ ) knowing that the 2

weighting function w(t ) = e −t . Therefore, we will have:

∆π t = ∑lj =1 α j H

j −1

(t ) + β (L )u t

+ δ (L )∆π t −1 + γ (L )X t + ε t

and the estimation of the varying NAIRU in function of time is:

− ∑ j =1 α j H l

ˆ*

u =

j −1

(t )

βˆ (1)

The selection of the order of the polynomial series of Hermite is given by Mallows Cp. The current model chooses series of Hermite in the order of 6. The Hermite polynomials, in particular, are orthogonal and generally have properties making them flexible forms1. To assess our estimation, and according to Sekhon (2001), it is preferable to use ut −1 and X t −1 instead of u t and X t . Thus, the estimated equation will have the following form:

∆π t = ∑lj =1 α j H

j −1

(t ) + β (L )ut −1 + δ (L )∆π t −1 + γ (L )X t −1 + ε t

1

We use Hermite polynomial to estimate the NAIRU. The series of Hermite is used to describe a groove cube. The polynomials of Hermite form one family of orthogonal polynomials. The matrix of variance-covariance is well conditioned. The polynomials of Hermite must be defined for j > 0 by the following recurrence relation: H j +1 (t ) = 2tH j (t ) − 2 jH j −1 (t ) , where j is the order and

H 0 (t ) = 1 and H 1 (t ) = 2t . The argument « t » has to be centred around 0 because polynomials are

orthogonal

on

the

2

symmetrical

+∞

interval

(− ∞;+∞ )

(t ) = e−t , ∫−∞ H j (t )H k (t )e −t dt =2 k k!π 0.5δ jk

function w Hermite

polynomial

H 4 (t ) = 16t 4 − 48t 2 − 12 .

is

2

listed

below.

knowing

the

weighting

; j , k ≥ 0 . A number limited of

H 2 (t ) = 4t 2 − 2 ;

H 3 (t ) = 8t 3 − 12t ;

41


2.2. Capacity Utilization and NAIRU However, to get a certain relationship between inflation and the CU, an extraordinary supposition has to be made on the formation of anticipated inflation. Specifically, this rate, *

noted π , is equal to a pondered average of a large number of lags of inflation, with the sum of weightings equal to the unit. To see if this supposition is necessary, we consider an empirical model of Phillips curve in which a set of inflation lags is going to represent the anticipated inflation: n

π t = b0 + b1ut + ∑ b2iπ t −i + ε t i =1

where π and u represents, respectively, the level of inflation and the unemployment rate, n

and

π t* = E (π t ) = ∑ b2iπ t − i . This situation is represented in Figure 1 where the i =1

inflation rate is equal to the anticipated one. Figure 1: Relation Inflation and Unemployment Inflation Rate %

10 5

π * = 10%

0

π * = 5% π * = 0% NAIRU

Unemployment Rate

n

In the ideal case where

∑b

2i

= 1 , we have:

i =1 n −1



i



π t − π t −1 = b0 + b1ut + ∑ δ i (π t −i − π t −i −1 ) + ε t , where δ i =  ∑ b2 j  − 1 . j =1 i =1 



In the long run, this equation has the form of Figure 2 when the change of retarded inflation is equal to zero.

42


Figure 2: Relation Inflation and CU Inflation Rate %

π * = 10% π * = 5%

π * = 0% 10 5

0

Natural rate of CU CU The NAIRU and the NAICU are special cases of natural rate where the anticipated supposition described above is invoked. Under this supposition, the result is a positive relation between the change of the CU and inflation, described in Figure 3. Figure 3: Relation Inflation and NAICU Inflation Rate %

0

NAICU

CU


43

If this assumption is not taken into account, then there will be on average instability in this relation. To examine the relation Inflation-CU, more precisely, we begin with a series of regression under the following form: n

∆π t = π t − π t −1 = b0 + b1CU t −1 + ∑ b2i π t −i + ε t i =1

where

∆π t is the variation of inflation between t and t-1 and b0, b1 and b2i are parameters

to estimate. Empirically, it is necessary to show that the lag of the CU is significant and of positive indicating that a high CU leads to an increase of inflation. Thus, the magnitude of the CU varies with data used. In general, a CU higher than a NAICU will lead to an increase of inflation. Following the excess of aggregate demand, firms tend to acquire high production costs in order to increase production. Firms are forced to non-efficient factors of production, inexperienced employees and non-performing equipment, entailing some elevated costs of production under the form of high selling prices. It is equivalent to say that the process of firm production presents some short-run decreasing outputs of scale. To study the relationship between the CU and inflation, two alternative specifications of Phillips curve were adopted. Firstly, a price standard model, which supposes that inflation follows a unit root process, is used to estimate Phillips curve in the short-run. Secondly, to test the magnitude of the results, the unit root assumption is released and a specification of Phillips curve is estimated. Therefore, the question is: is the CU going to remain an efficient instrument to anticipate inflation? Traditionally, the objective of using the CU gap instead of the unemployment rate, as a measurement of the inflationary pressure, was the NAICU which should be stable on a given level. There are two main ways where the CU can affect inflation. Firstly, high CU are associated with high costs (decreasing return of scale in the short-run), thus forcing the prices to increase (when the offer is relatively rare as compared to the demand. Consequently, the producers increase the prices without any serious loss in the sales). These forces affect the prices of goods directly, but they must be important for the foreign price index. Secondly, the constraints of the quantity of the capacities encourage the firms to invest in new plans and equipments, by simulating the economic expansion and the pressures of capacity which gives rise to other inflationary pressures. The model used to test the validity of above hypotheses is similar to that of Emery and Chang (1997) and Baylor (2001), which represents a modified version of the price model used for the assessment of Phillips curve (McElhattan, 1985).


44

The starting point is the canonical Phillips curve:

π t = b0 + b1CU t + b2π t* + ε t We assume that anticipated inflation is equal to the pondered average of lagged inflation and the sum of weights is equal to the unit (anticipation follows a unit root process). This implies that there is a stable and permanent relation between the changes of inflation and the CU. In rewriting equation (12), we obtain: n

π t = b0 + b1CU t + ∑ b2i π t −i + ε t , where i =1

n

∑b

2i

= 1 , we will be able to rewrite

i =1

equation (13) as: n −1   i π t − π t −1 = b0 + b1CU t + ∑ δ i (π t −i − π t −i −1 ) + ε t , where δ i =  ∑ b2 j  − 1 .   j =1 i =1   Consequently, the change of inflation is connected to the CU and to the previous changes in inflation. Similarly, when the lags of inflation are null ( π t −i − π t −i −1 = 0 ), this − bˆ0 equation is an ascending straight line that crosses the horizontal axis at NAICU = . bˆ1 The measures of excess of demand and the previous inflations are not the only determinants of current inflation. The different exogenous events, such as the offer shocks, affect prices significantly. The changes of oil prices and real exchange rate can affect prices directly or indirectly. In order to avoid a bad specification, it is evident to introduce variables of control. Thus, equation (14) will be added up by a vector of control variable in the case of the offer shocks. Thus, we reach a final specification of the model:

∆π t = b0 + b1CU t +

n −1

m

i =1

j =1

∑ δ i ∆π t −i + ∑ λ j z jt + ε t

where the zj are variables of control. 2.3. Effect of Capacity Utilization and Unemployment on Inflation

Phillips curve is the best known relation in specifying inflation as a function of certain measures of insufficiency of productive resources, namely unemployment rate, gap of unemployment and gap of production. The traditional Phillips curve employs the unemployment rate and in certain cases the lag of inflation. Thus, inflation, π , is defined as the lag function of the difference between the natural unemployment rate and the current unemployment rate, υ t = u t − u e , and the lag of inflation, π t −1 :

45


π t = b0 + b1υ t −1 + b2π t −1 + ψ t This equation is similar to Phillips curve estimated by Finn (1995) and Belton and Cebula (2000). In addition to the unemployment rate and the lag of inflation, we can take into account, explicitly, the aggregated offer shocks and the monetary policy. The Offer shocks are the consequences of the oil shocks and the openness to the developed economies. The monetary shocks are important because they play a major role in information to make a good anticipation of inflation. The two oil shocks of the 70s accelerated inflation over a period of 10 years. Inflation rose above 8% in 1974, just after the first oil shock, and 14 % by the end of 1982, after the second oil shock that of 1979. In fact, oil shocks are characterized by an increase of oil prices, leading to a rise of inflation. The openness of the Tunisian economy to the developed countries is another factor that can influence the domestic economy. The increase of inflation in the foreign countries can, consequently, lead to an increase of home inflation following the demand for domestic goods. On the other hand, a large openness to foreign countries can help population meet its needs and solve the problem of shortage of production resources; as a result, there will be a decline in domestic inflation. Since the monetary policy is in the hands of the monetary authorities (Tunisian Central Bank), its goals and attempts are unknown. Hence, monetary shocks are likely to happen. To examine the relationship between the CU and inflation, it is necessary to put them together in the Phillips curve. Equation (17) shows that inflation is a function of the offer and monetary policy shocks, the lag of inflation, the CU or/and the gap of unemployment rate.

(

)

(

)

π t = α0 + α1COt + α 2 IPt + α 3 mtA−1 − m*t −1 + α 4π t −1 + α 5 ut −1 − u* + α6 CU t −1 + ε t where π t is inflation at period t, COt is a Proxy of the two oil shocks of the 70s, IPt is a

(

)

Proxy of the non energizing import shocks. mtA−1 − m*t −1 is a lag of the difference

(

)

between the current currency and the currency announced as objective, u 1 = u t −1 − u * is a lag of the gap of unemployment rate, π t−1 is a lag of inflation, and CU t −1 is a lag of capacity utilization. 3. Empirical Analysis

The study of the relationship between inflation and the CU covers the annual data of the following macroeconomic variables from 1970 to 2006: CU, capacity utilization rate


46

estimated according to constant return to scale Translog cost function2; CPI, consumer price index as a measure of inflation; u, unemployment rate; g, monetary growth rate expressed by M2. To start our empirical study, we have to test each variable for its stationary condition. The Used models are based on the supposition of the stationary condition of all the variables in the regression. To identify the order of integration of our variables, we used the test of Augmented Dickey and Fuller (1979) and that Phillips and Perron (1988). These tests presuppose recognizing the model structuring the chronicle while in reality they don’t. Having used the annual observations (37 observations), we could experiment until k_max = 3 of lags on the first difference of every variable. In summary, the unit root results are presented in Table 1. Table 1: Unit Root Test in Level and First Difference3 In Level Variables

k_max

CU CPI u g

1 2 3 1

ADF Stat. -3,67b -1,55 -2.75c -4.89a

k 2;T 1;_ 3;C 0;T

PP Adjust 0,01 -3,61b -1.08 -4.89a

L.B 1;_ 3;T 3;_ 3;T

In First Difference ADF PP Stat. k Adjust L.B -7,73a 1;T -7,61a 0;T -9,26a 0;T -9,6a 3;T -3.68b 0;T -3.68b 3;T a a -8.21 0;T -9.55 3;T

As indicated in Table 1, most of the variables are integrated of the order "1" with the exception of the monetary growth rate (g). The non-stationary condition is originally determinist and stochastic for all the variables, specifically for the CU. The first difference of each of these variables is however, stationary. 3.1. Estimation of the NAIRU

In reference to Sekhon (2001) and Ball and Mankiw (2002), we are going to develop a simple model permitting to estimate the NAIRU of the Tunisian economy. The NAIRU model, as a standard model of inflation, is based on the anticipation of augmented Phillips curve. 2

See PhD Helali K. (2010). k represents the optimal number of lags where the statistics_t in the regression is significant, L.B represents the bandwidth by using Bartlett kernel in the test of PP, C represents a significant constant, T represents a constant and a significant trend and (_) represents neither tendency nor constant. (a), (b) and (c) represent respectively the significance at levels off 1%, 5% and 10%. k_max is selected according to the Schwarz information criterion (SBC). 3

47


The tests of the stationary condition of the variables of inflation model (4) show the stationary condition of both the CPI and g, while unemployment rate "u" is not stationary in level. To solve this problem of the integration order, we employ the ARDL approach (Autoregressive Distributed Lag Model) of cointegration. This approach is a technique of the relatively recent econometrics developed by Pesaran and al. (2001) to estimate the long-run relationship between variables. This approach tests the relationship of cointegration without requiring the same order of integration of all variables. Figure 4: Evolution of Unemployment with inflation 20%

15%

19%

13%

18% 11%

17% 16%

9%

15% 7%

14% 13%

5%

12% 3%

11%

Unemployment

2006

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

1974

1972

1% 1970

10%

Inflation

The ARDL model is probably the most largely employed model in estimating the relationships in the context of the temporal series. In this work, we wonder whether inflation and the CU are co-integrated by using the new procedure "Bounds tests" to analyze the degree of relationship with the ARDL structure. The added value of the procedure employed in the present contribution allows to test cointegration with certainty when it is not known and the regressions are purely I(0), purely I(1) or mutually cointegrated. The ARDL approach of cointegration implies an estimation of the version of conditional error correction of the ARDL model for inflation and the CU. This representation is formulated as follows: p

∆y t = α 0 + ∑ λi ∆y t −i + i =1

q

q

j =0

j =0

∑ β j ∆xt − j + ∑ γ j ∆z t − j + δ 1 y t −1 + δ 2 xt −1 + δ 3 z t −1 + ϕ t

where y t = ∆π t , xt = u t et z t = ∆g t .

48


Table 2: shows the negative effect of unemployment rate on the variation of inflation in Tunisia measured by the CPI. The inflation equation takes its efficiency from the lag of inflation and more precisely from an ARDL(1,0,0). The chosen (g) Proxy has not shown its efficiency on inflation as its was shown previously. The long-term estimation permits to deduce some punctual estimation of the NAIRU for the Tunisian case. Table 2: Estimation of the Phillips curve "Short-run Results"4 ( y t = ∆π t ) 1 Statistics F*** 1.403 SBC Criteria ARDL Model (1 , 0 , 0)

Order of lags AIC Criteria

∆ (∆π )t

11.51 1.717

Variables

p-values

∆ (∆π )t

Variables

p-values

∆(∆π )t−1

0.036

0.793

ut −1

-0.338

0.046

∆u t ∆(∆g )t

-0.44

0.299

∆g t −1

0.135

0.141

0.028

0.551

Constant

4.733

0.056

∆π t−1

-1.498

0,000

2

0,79 DW 2.28 R The balanced unemployment rate, which does not permit to accelerate inflation, is evaluated by the NAIRU at a rate of 14% with an interval of confidence of 95% of significance and equal to [12.6 ; 15.4]%. Table 3: Estimation of NAIRU "Long-term Results" ( y t = ∆π t )

Constant ut

Coefficient 3.161 -0.226

p-values 0.031 0.024

NAIRU Standard Deviation

∆g t

13.99 % 0.72

0.090 0.163 CI at 95% 12.6 - 15.4 The estimated approach can change in case where the NAIRU varies with time. By referring to equations (6) and (7), we can add the Hermite polynomials in the estimation above, and then we will have three important remarks. Firstly, we choose three polynomials (H , H¹ and H²) since the polynomials of an order superior to 2 have too weak and non-significant effects. Secondly, the number of lags passes from 1 to 3 for the CPI and so we choose the ARDL(3,0,0) approach. Thirdly, if we change variables ut and ∆g t by ut-1 and ∆g t −1 , the number of lags passes to 2 and then we choose the ARDL(2,0,0) approach for the first two Hermite polynomials. 4

*** means that the statistic F is above the boundary-mark superior to 99% for k=2.

49


The estimations of unemployment rate of Table. 4, with lags (column 2) and without lags (column 3); do not show a strong divergence. A lag of unemployment rate is more significant than variable in level. Similarly, the Proxy chooses ( ∆g t ) has not shown its efficiency on inflation. From long run estimations, we estimate the variant NAIRU. Table 4: Estimation of Variant Phillips Curve "Short-run Results"5 ( y t = ∆π t ) ARDL Model

∆(∆π )t−1 ∆ (∆π )t− 2 ∆(∆π )t−3 ∆u t ∆(∆g )t ∆π t−1 ut −1 Order of lags Statistics F***

(3 , 0 , 0) 1.117 (0.000) 0.721 (0.000) 0.225 (0.099) -0.695 (0.332) 0.004 (0.938) -2.784 (0.000) -0.781 (0.117) 3 10.92

(2 , 0 , 0) 0.641 (0.013) 0.329 (0.042)

ARDL Model

(3 , 0 , 0) 0.064 (0.441) 9.048 (0.010) 0.074 (0.699) -0.6e-3 (0.733)

(2 , 0 , 0) 0.161 (0.092) 6.356 (0.002) 0.002 (0.840)

R2

0.73

0.75

DW

1.99

2.23

AIC Criteria SBC Criteria

1.446 1.950

1.339 1.747

∆g t −1 H0

_

H1

-0.536 (0.052) 0.084 (0.158) -2.226 (0.000) -0.465 (0.001) 2a 10.98

H2

_

Long term results show negative and significant effect of unemployment rate on inflation. With one lag the result is nearer to previous estimation. We can distinguish this observe the Figure. 5 that shows the evolution of the two estimations.

5

*** means that the statistic F is above the boundary-mark superior to 99% for k=2. a: Estimation with one lag of u t −1 and ∆g t −1 . Value between parentheses is p-values.

50


Table 5: Estimation of Variant NAIRU "Long-run Results"6 ( y t = ∆π t ) Constant b

ut

3.821 (0.000) -0.280 (0.078)

2.891 (0.000)

NAIRU

13.62 %

13.82 %

_

Standard Deviation

0.77

0.10

CI at 95%

12.1 – 15.1

13.8 – 14.2

ut −1

_

-0.209 (0.000)

∆g t

0.023 (0.441)

_

∆g t − 1

_

0.072 (0.085)

The NAIRU estimated points, represented above in the Figure, show weak growth of about 2.5% between 1970 and 2006 In fact, if u < NAIRU for a weak period, the inflationary pressures increase so that inflation tends to accelerate. It is the situation that characterized the 1975-1985 phase where we observed a strong rise of inflation (14%) accompanied by a financial crisis explaining the escalation of the 1983-1985. Simultaneously, we record a strong decrease of the CU reaching a minimum of 43%. Figure 5: Evolution of NAIRU in level and with one lag 15,0% 14,5% 14,0% 13,5% 13,0% 12,5% 12,0% 11,5% 11,0%

NAIRU

2006

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

1974

1972

1970

10,5%

NAIRU1

However, if we have u > NAIRU, the inflationary pressures decrease and inflation tends to slow down (there is a deflation phenomenon). This phase characterized the period where the Tunisian economy had recovered. The structural program of 1987-1988 could free up 6

b: an average value of the series α 0 = ∑ I α j H j =1

j −1

(t ) .

51


most prices of the financial sector, the thing which helped regain a competitive environment. Therefore, the economy comes back to a moderate inflation pace permitting the pursue of a satisfactory macro-economic policy. This is justified by the reuse of capacity utilization. Nevertheless, the Tunisian economy could not reduce unemployment. Figure 6: Evolution of Unemployment Rate with NAIRU 20% 19% 18% 17% 16% 15% 14% 13% 12% 11%

Unemployment

NAIRU

NAIRU inf

2006

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

1974

1972

1970

10%

NAIRU sup

Finally, if u = NAIRU, inflation tends to remain steady, unless there are exogenous shocks. Among these shocks, we mention the second gulf war of 1991 that increased inflation to 8.2% and lowered the CU until 50% whereas the unemployment rate was found to be surrounded by the confidence interval of the NAIRU. Figure 7: Evolution of CU with Unemployment Rate

Unemployment

CU

2006

2004

2002

2000

1998

1996

1994

1992

40% 1990

45%

10% 1988

50%

11% 1986

55%

12%

1984

60%

13%

1982

65%

14%

1980

70%

15%

1978

75%

16%

1976

80%

17%

1974

85%

18%

1972

90%

19%

1970

20%

52


Although economists have discussed the relative ability of the unemployment rate and the CU in explaining inflation extensively, few formal empirical researches were carried out in order to explain it. Finn (1995 and 1996), in particular, divides the CU in two components, one is high and the other is low in relation to the balance rate. Actually, she shows that only high rates have a significant effect on inflationary pressures. 3.2. NAIRU and CU Relationship

To check these relationships empirically for the case of the Tunisian economy, it is necessary to combine inflation and the CU in the same Phillips curve. Equation (17) shows that inflation is a function of the offer shocks, the monetary policy shocks, the lag of inflation, and the CU or/and the gap of unemployment rate. To examine the strength of the Inflation-CU relationship with more precision, it is necessary to use the PPI "Production Price Index" to make a comparison with the CPI. The test of the position of the used variables accepts the null hypothesis for CO and g, and rejects it for the others. So we apply the ARDL approach of the following shape: p

∆y t = α 0 + ∑ λi ∆y t −i + i =1

q

q

q

q

j =0

j =0

j =0

j =0

∑ β j ∆xt − j + ∑ γ j ∆z t − j + ∑ ρ j ∆wt − j + ∑ φ j ∆mt − j

+ δ 1 y t −1 + δ 2 xt −1 + δ 3 z t −1 + δ 4 wt −1 + δ 5 mt −1 + ϕ t where y t = π t , xt = (CU t −1 )t , z t = (u1 )t , wt = COt et mt = IPt . Starting from Table. 6, we can draw three essential results. Firstly, we distinguish a difference of order for both measures of inflation with the other variables. Secondly, the CU and unemployment corrected by the NAIRU showed opposite and significant effects on inflation. Thirdly, the Proxy CO did not show a strong influence unlike the IP, which shows the fundamental effect of imports on the fall of domestic prices. In the long run, we clearly observe the negative effect of the CU and a positive effect of unemployment. The significance both factors allows us to combine these two indicators of demand in the same equation. As for the inflation measured by the CPI, the Proxy of nonenergy import shocks proves its significant effect. In fact, the CU did not verify the theoretical concepts of a positive effect on inflation. Perversely, we observe a decrease of inflation, which indicates that the CU in Tunisia is not a source of inflationary pressures. Inflation by consumption is influenced too much by the import of non-energy goods because the oil crisis of the 1970s didn't affect the increase of inflation. Besides, unemployment can be a source of inflationary pressures.

53


Table 6: Effect of CU and Unemployment on Inflation "Short-run Results"7 (2 , 1 , 0 , 0 , 0 )

(1 , 0 , 0 , 0 , 0 )

∆CPI t

∆PPI t

Order of lags

2

1

∆COt

Statistics F**

4.21

5.69

∆IPt

AIC Criteria

0.782

1.689

π t−1

SBC Criteria

1.366

2.178

(CU t −1 )t −1

0.124 (0.457) 0.196 (0.213) -0.167 (0.011) -0.149 (0.049) 0.228 (0.546)

0.231 (0.259)

(u1 )t −1

_

COt −1

-0.438 (0.001)

IPt −1

ARDL Model

∆π t−1 ∆π t− 2 ∆(CU t −1 )t ∆(CU t −1 )t −1

∆ (u1 )t

_ 0.755 (0.176)

ARDL Model

(2, 1, 0, 0, 0 ) (1,0 ,0 ,0, 0)

∆CPI t

∆PPI t

0.009 (0.116) -5.49 (0.058) -2.784 (0.000) -0.207 (0.000) 0.612 (0.012) 0.018 (0.107) 6.305 (0.000) 9.579 (0.002)

-0.002 (0.777) -3.053 (0.289) -2.226 (0.000) -0.211 (0.000) 1.547 (0.001) 0.026 (0.036) 13.57 (0.272) 2.393 (0.272)

R2

0.62

0.65

DW

2.19

2.20

Constant

Unlike the CPI, the PPI equation shows a significant effect of all Proxy's. Their effects entail an increase of inflation by production. Similarly, we observe a negative effect of the CU, which justifies the conclusion that the CU is not an efficient instrument of inflationary pressures for the Tunisian case. The first lag of the CU and unemployment show opposite effects on inflation, which contradicts the previous theoretical concepts of complementary. These two measures of unbalance related to the demand can be substitutes in the inflation equation.

7

** means that the statistic F is above the boundary-mark superior to 95% for k=4. Value between parentheses is p-values.

54


Table 7: Effect Of CU and Unemployment on Inflation "Long-Run Results"

Constant

CU t −1

(u1 )t

CPI t

PPI t

9.890 (0.000) -0.214 (0.000) 0.632 (0.029)

2.007 (0.272) -0.177 (0.000) 1.297 (0.000)

COt IPt

CPI t

PPI t

0.018 (0.117) 6.509 (0.000)

0.022 (0.001) 11.38 (0.000)

The conclusion reached in this survey is that the CU in Tunisia does not present an efficient indicator of inflationary pressures for different reasons. First, in the previous years, the Tunisian economy experienced a stability of inflation, the thing which does not permit to distinguish the different fluctuations of the inflation trajectory measured by the consumer price index or production price index. Secondly, the oil price shocks or the financial crises happening during the period of survey have been solved either by the economic growth of the 70s or by the plan of structural adjustment of 1987-88. The problems of the Tunisian economy result in the bad management of the use of the production resources without strongly influencing the financial evolution of the economy. 4. Conclusion

This paper states that the relationship is stable and the capacity utilization rate remains an efficient estimator of the future change of inflation. Specifically, we have used a simple regression of an ordinary least square, based on the ARDL models, to show that, based on different samples; the Non Accelerating Inflation Capacity Utilization is dramatically constant in the range of 65% for Tunisia. Besides, we have underlined the similarity between the analyses using the NAIRU concept and those using the capacity utilization rate (CU). The best means of showing this similarity is by the replacement of the unemployment rate with the CU in a simple model of anticipated and increased Phillips curve. These curves express a negative exchange between the levels of the inflation rate and that of unemployment for a given level of anticipated inflation. In addition, the statistical and economic experiences showed that the short-run CU is not an efficient indicator of inflationary pressures for the Tunisian case, however, in the long run, we notice a positive effect of this relationship. This could be explained by the fact that the economy is characterized by fast gains of productivity. At this level, the economy follows major structural changes, such as machine incorporation, the use of new computers and technologies of telecommunication, and high levels of investment. The


55

extreme version of this argument is that capacity is unlikely to be a constraint of a short run economic growth because of the improvements of the industrial productivity. In this case, there is no virtual speed limit for the economy so as to grow with a steady inflation. The weak estimated relationship in the Tunisian economy between the CU and inflation is extensively due to the absence of capacity utilization as an indicator of inflation if we know that it is calculated only for the industries of production of goods while ignoring the growth of service industry. References

Ball, L. and Mankiw, N., 2002, The NAIRU in theory and practice, NBER Working Paper 8940. Baylor, M, 2001, Capacity Utilization and Inflation: Is Statistics Canada's Measure an Appropriate Indicator of Inflationary Pressures? Department of Finance Working Paper 2001-06. Belton, Willie J, Jr and Cebula, Richard J, 2000, Capacity Utilization Rates and Unemployment Rates: Are They Complements or Substitutes in Warning about Future Inflation? Applied Economics, Taylor and Francis Journals, 32, 12, 1521-32. Berndt, R. Ernst and Morrison, Catherine J., 1981, Capacity utilization measures: Underlying Economic Theory and an Alternative Approach, American Economic Review, 71, 48-52. Corrado, C. and J. Mattey, 1997, Capacity Utilization, Journal of Economic Perspectives, 11, 1, 151-167. Dickey, D. A. and Fuller W. A., 1979, Distribution of the Estimators for Autoregressive Time Series with a Unit Root, Journal of the American Statistical Association, 74, 417431. Emery, Kenneth M. and Chih-Ping Chang, 1997, Is There a Stable Relationship between Capacity Utilization and Inflation, Federal Reserve Bank of Dallas, Economic Review, First Quarter, 21-20. Finn, Mary G., 1995, Is "high" capacity utilization inflationary? Federal Reserve Bank of Richmond, Economic Quarterly, 81, 1-16. Finn, Mary G., 1996, A theory of the capacity utilization/inflation relationship. Federal Reserve Bank of Richmond, Economic Quarterly, 82, 3, 67-86.


56

Garner, Alan C., 1994, Capacity utilization and U.S. inflation, Federal Reserve Bank of Kansas City, Economic Review, Fourth Quarter, 1-21. Gordon, Robert J., 1989, Assessing the Federal Reserve's Measures of Capacity and Utilization: Comment, Brookings Papers on Economic Activity, 2, 226-35. Gordon, Robert J., 1995, Estimating the NAIRU as a Time-Varying Parameter, Paper presented at the Meeting of Academic Consultants at the Congressional Budget Office, Washington, D. C., November. Helali, Kamel, 2010, Taux d'utilisation des capacités de production et inflation en Tunisie : analyse théorique et étude économétrique, Thèse de Doctorat, FSEG Sfax. McElhattan, Rose, 1978, Estimating a Stable-Inflation Capacity Utilization Rate, Federal Reserve Bank of San Francisco, Economic Review, 20-30. McElhattan, Rose, 1985, Inflation, supply shocks and the stable-inflation rate of capacity utilization, Federal Reserve Bank of San Francisco, Economic Review, Winter, 45-63. Nahuis, Niek J., 2003, An alternative demand indicator: the non-accelerating inflation rate of capacity utilization, Applied Economics, Taylor and Francis Journals, 35, 11, 13391344, July. Pesaran, M. H., Shin, Y. and R. J. Smith, 2001, Bounds Testing Approaches to the Analysis of Level Relationships, Journal of Applied Econometrics, vol. 16, pp. 289-326. Phillips, P. C. B. and Perron P., 1988, Testing for a Unit Root in Time Series Regression, Biometrika, 75, 335-346. Rennison, A., 2003, Comparing Alternative Output-Gap Estimators: A Monte Carlo Approach, Document de travail no. 2003-8, Banque du Canada, Ottawa. Sekhon, J. S., 2001, Estimation of the Non-accelerating Inflation Rate of Unemployment, Center of Basic Research in the Social Science, Harvard University. Shapiro, Matthew, 1989, Assessing the Federal Reserve's measures of capacity and utilization, Brookings Papers on Economic Activity, 1, 181-225. Staiger, Douglas, James H. Stock, and Mark W. Watson, 1997, The NAIRU, Unemployment and Monetary Policy, The Journal of Economic Perspectives, 11, 33-50. Tootell, Geoffrey M. B., 1994, Restructuring, the NAIRU, and the Phillips Curve, Federal Reserve Bank of Boston, New England Economic Review, September/October, 31-44. Weiner, Stuart E., 1993, New Estimates of the Natural Rate of Unemployment, Federal Reserve Bank of Kansas City, Economic Review, 78, 4, 53-69.