A. Convergence of Inflation Rates

Let p_it denote the logarithm of the consumer price index (CPI) at time t in country i, where t = 1,…,T and i = 1,…,N. The deviation of the price level between country i and j can then be written as:

implying that the inflation differential between country i and j is given by:

Before proceeding to evaluate the factors that determine short-run deviations of inflation between countries i and j, we must first establish that pairwise inflation rates in the GCC region are in fact converging, i.e. stationary, otherwise any inference on the estimations of Δx_ijt on its short-run determinants would be invalid. Unit root tests are used to determine whether pairwise inflation rates are in the process of converging. A model testing whether the inflation rates are converging will be asymptotically stationary, satisfying the condition that:

where ΔX_ijt denotes the current and past observations. Convergence is said to be absolute if α_ij = 0, otherwise it is conditional or relative, see Durlauf and Quah (1999). A very simple convergence model is the AR(1) process:

where $${\xi }_{ijt}$$ is a martingale difference innovation and Δx_ij0 is a fixed initial condition. Writing equation (4) in error correction form:

where $${\gamma }_{ij}=\left(1-{\varphi }_{ij}\right){\alpha }_{ij}$$, we can see that after allowing for a permanent difference, α_ij, the expected growth rate in the price level in the current period is a negative fraction of the gap in the two countries. A test for convergence can then be done by performing a unit root test, in other words $$H_0:\varphi =1$$ against the alternative $$H_1:\varphi <1$$. For a richer dynamic we will consider an AR(p) process:

on which an Augmented Dickey-Fuller (ADF) test is based. While in many cases we would consider investigating both absolute and conditional convergence, it is implausible that there should exist permanent inflation differentials across the GCC countries. We test for absolute convergence by verifying whether pairwise inflation differentials, Δx_ijt, in the GCC are converging without there being a permanent and persistent influence on the inflation differentials. Absolute convergence is also ultimately the aim in a monetary union, and forms the basis for a convergence criterion on inflation rates before a country can join a monetary union.

To investigate the convergence properties of inflation rates in the GCC, we obtain monthly CPI data from the International Monetary Fund (2010b), between January 1991 and June 2010. The pairwise combinations are ordered by countries according to their GDP weights in the GCC area and are listed in Table 1. Table 2 provides a definition of the variables used in this paper as well as their sources. After constructing pairwise inflation differentials, unit root tests were performed by estimating equation (6). The results of the ADF tests reported in Table 3 show that for all of the 15 pairwise combinations we can reject the null of a unit root when investigating absolute convergence, α_ij = 0. Moreover, all but two pairwise combinations (Saudi Arabia with the U.A.E. and Qatar with Oman) are also stationary when we allow for α_ij ≠ 0. Given that for these two combinations we cannot reject the null of non-stationarity we also perform a stationarity test as proposed in Kwiatkowski et al. (1992), denoted as KPSS and reported in Table 3, which shows that for these two particular combinations the more strict null of stationarity cannot be rejected. Therefore, the empirical evidence points to absolute convergence in the GCC pairwise inflation differentials over the period 1991 to 2010.

Table 3.

Unit Root and Stationarity Tests on Pairwise Inflation Differentials, 1991M1–2010M6

Notes: ADF refers to the Augmented Dickey—Fuller test while KPSS is the Kwiatkowski-Phillips-Schmidt-Shin test. A ^* indicates that the null hypothesis cannot be rejected. Authors’ calculations based on data from International Monetary Fund (2010b).

Table 3.

Unit Root and Stationarity Tests on Pairwise Inflation Differentials, 1991M1–2010M6

Pairwise Combinations				ADF, no intercept		ADF, with intercept		KPSS, with intercept
				Statistic	Reject	Statistic	Reject	Statistic	Reject
Saudi Arabia – Bahrain				-16.56	1%	-16.62	1%	0.403	10%
Saudi Arabia – Kuwait				-10.52	1%	-10.57	1%	0.220	*
Saudi Arabia – Oman				-16.23	1%	-16.22	1%	0.175	*
Saudi Arabia – Qatar				-1.98	5%	-2.67	10%	0.264	*
Saudi Arabia – UAE				-2.06	5%	-2.23	*	0.229	*
			UAE – Bahrain	--2.80	1%	-3.67	1%	0.123	*
			UAE – Kuwait	--5.00	1%	-9.44	1%	0.327	*
			UAE – Oman	--2.00	5%	-15.90	1%	0.590	5%
			UAE – Qatar	--4.00	1%	-4.04	1%	0.259	*
	Kuwait – Bahrain			-9.56	1%	-9.74	1%	0.367	10%
	Kuwait – Oman			--8.90	1%	-8.99	1%	0.150	*
	Kuwait – Qatar			-5.07	1%	-5.42	1%	0.192	*
			Qatar – Bahrain	-2.58	1%	-2.97	5%	0.199	*
			Qatar – Oman	-2.17	5%	-2.43	*	0.262	*
		Oman – Bahrain		-16.06	1%	-16.03	1%	0.579	5%

Notes: ADF refers to the Augmented Dickey—Fuller test while KPSS is the Kwiatkowski-Phillips-Schmidt-Shin test. A ^* indicates that the null hypothesis cannot be rejected. Authors’ calculations based on data from International Monetary Fund (2010b).

View Table

Table 3.

Unit Root and Stationarity Tests on Pairwise Inflation Differentials, 1991M1–2010M6

Pairwise Combinations				ADF, no intercept		ADF, with intercept		KPSS, with intercept
				Statistic	Reject	Statistic	Reject	Statistic	Reject
Saudi Arabia – Bahrain				-16.56	1%	-16.62	1%	0.403	10%
Saudi Arabia – Kuwait				-10.52	1%	-10.57	1%	0.220	*
Saudi Arabia – Oman				-16.23	1%	-16.22	1%	0.175	*
Saudi Arabia – Qatar				-1.98	5%	-2.67	10%	0.264	*
Saudi Arabia – UAE				-2.06	5%	-2.23	*	0.229	*
			UAE – Bahrain	--2.80	1%	-3.67	1%	0.123	*
			UAE – Kuwait	--5.00	1%	-9.44	1%	0.327	*
			UAE – Oman	--2.00	5%	-15.90	1%	0.590	5%
			UAE – Qatar	--4.00	1%	-4.04	1%	0.259	*
	Kuwait – Bahrain			-9.56	1%	-9.74	1%	0.367	10%
	Kuwait – Oman			--8.90	1%	-8.99	1%	0.150	*
	Kuwait – Qatar			-5.07	1%	-5.42	1%	0.192	*
			Qatar – Bahrain	-2.58	1%	-2.97	5%	0.199	*
			Qatar – Oman	-2.17	5%	-2.43	*	0.262	*
		Oman – Bahrain		-16.06	1%	-16.03	1%	0.579	5%

Notes: ADF refers to the Augmented Dickey—Fuller test while KPSS is the Kwiatkowski-Phillips-Schmidt-Shin test. A ^* indicates that the null hypothesis cannot be rejected. Authors’ calculations based on data from International Monetary Fund (2010b).

B. Econometric Model

A fairly general specification for inflation in country i can be written as:

for i = 1,2,…,N and t = 1,2,…,T

where $$p_{it-1}^{*}$$ is the long-run equilibrium price level in country i, α_i is the country-specific effect, and z_it includes those macroeconomic variables that exert short-term influence on the inflation process in country i. We assume that the error term, u_it, has the following multi-factor error structure:

where f_t is a vector of unobserved common shocks, which can be stationary or nonstationary, see Kapetanios et al. (2011). The impacts of these common components on each country are governed by the idiosyncratic loadings in θ_i. The individual-specific errors, ε_it, are distributed independently across i and t; they are not correlated with the unobserved common factors or the regressors; and they have zero mean, variance greater than zero, and finite fourth moments.

Absolute convergence then requires that:

and under the assumption that θ_if_t ~ I(1) it is also required that:

See Pesaran (2007) for an extensive explanation of these conditions when applied to cross-country output convergence. Thus, given convergence in inflation rates in the long-run, any short-run deviations of inflation across the countries are explained by deviations of the variables in z_it from z_jt.

Using equation (7) together with the conditions in (9) and (10), we can write an equation for inflation differentials between countries i and j:

Given that the GCC countries are in the process of establishing a monetary union, already have a set of convergence criteria in place, have a common peg to the dollar, and have taken several steps as mentioned in Section I to move to this goal, we can assume that the GCC countries share a common long-run price level, $$p_{it}^{*}=p_{jt}^{*}$$, but we allow for short-run deviations in their price levels:

Our model for inflation differentials then requires that Δx_ijt is stationary, which we have shown to be the case, otherwise estimations of equation (12) are not valid. In fact, given that we have absolute convergence in inflation differentials in the GCC region, thus implying that the price levels are converging to the same level across the countries (see Table 3), the assumption of $$p_{it}^{*}=p_{jt}^{*}$$ seems justified.

The main difference between our model and most others used in the literature is that we look at the pairwise inflation differentials, using a similar approach as in Pesaran (2007) and Pesaran et al. (2009), as opposed to the deviation of each country from a monetary union benchmark. For instance, Honohan and Lane (2003) investigate a similar specification to ours for the initial twelve countries that joined the euro area, but assuming $$p_{it}^{*}=p_{jt}^{*}=p_t^{u*}$$ (in which $$p_t^{u*}$$ is the long-run union price level) they define inflation differentials as the deviation of country i from the Union:

where a superscript u defines the Union variables which is just the simple average of all the countries considered. They argue that the Union variables can be combined into a time dummy, δ_t, and so equation (13) can be rewitten as:

Although they estimate equation (14), it is unclear why the estimated coefficients of φ and β in equations (13) and (14) are identical and measure the same effects of differentials in macro variables on inflation differentials unless some strict assumptions are satisfied. Specifically it requires that the effect of unobservables which are captured in δ_t is identical for each of the countries, in other words δ_it = δ_t. This is quite a strong assumption which may not hold in reality given that countries in any monetary union are likely to be diverse. This is true even for the euro area, just comparing the German experience with that of the Greek or the Irish will underline the importance of a model in which δ_it should be allowed to be different in the cross-section. Note that the strong assumption of homogeneity can of course not be tested due to the fact that δ_it is unobserved.

Another general problem with estimating differentials as deviations from a union mean is that there is no particular reason why an equal weight of 1/N on the macro variables of the countries in our sample would apply, as it is completely plausible that some countries should be given a larger weight in the union. Thus an equation such as (14) seems more likely to be estimating individual country inflation equations, like that of equation (7) but allowing for time effects, as opposed to inflation differentials.³

Our pairwise approach on the other hand, equation (11), explicitly estimates φ and β as measuring the effects of differences in country i and j’s macro variables on explaining the different dynamics in the inflation rates for those countries. In addition, our approach does not depend on assigning an arbitrary weight on each country to generate Monetary Union macro variables, which is needed to estimate equation (13). Finally, estimation of equation (14) requires that Δp_it is stationary which in some cases might not be valid, see for instance Garratt et al. (2006) in which Δp_it is I(1) for the UK. Our approach on the other hand requires that the inflation differentials between the countries are stationary, which is shown to be valid in Table 3.

C. GMM Methodology

We begin with a basic specification that can nest much of the existing work on inflation differentials:

where all the variables are as defined in Section III.B. Traditional panel data estimators such as fixed and random effects are not consistent in the present context due to the inclusion of lagged dependent variables in our regressions, e.g. the initial level of price differentials. Moreover, these estimators are rendered inconsistent should the errors show either heteroscedasticity or serial correlation.

To correct for the bias created by lagged endogenous variables and the simultaneity of short-run inflation determinants, we use the GMM estimators developed for dynamic panel data models. Following Anderson and Hsiao (1982), and Arellano and Bond (1991), we take first-differences of equation (15), thus yielding:

The first-differences of equation (15) give the transformed error a moving-average structure that is correlated with the differenced lagged dependent variable. Assuming that the error term, (ε_it − ε_it), is not serially correlated and that the explanatory variables (z_it − z_jt) are weakly exogenous,⁴ the difference GMM estimator uses the following moment conditions:

However, in inflation regressions where the explanatory variables are persistent over time, lagged levels are often weak instruments for difference equations.⁵ To reduce the potential biases and imprecision associated with the GMM difference estimator, we employ a system estimator that in addition to the system of first differenced equations (16) also includes the original equation (15) in levels, but with the lagged differences of the endogenous variables as instruments (see Arellano and Bover (1995) and Blundell and Bond (1998)). Therefore, the additional moment conditions for the regression in levels are:

The moment conditions effectively give us T – 1 equations in first differences followed by T equations in levels. The solutions to these equations are then weighted by the inverse of a consistent estimate of the moment condition covariance matrix in a two-step method. To test the validity of the instruments and therefore the consistency of the GMM estimator, we consider two specification tests suggested by Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998). The first is a Hansen test of over-identifying restrictions, which tests the overall validity of the instruments and the second test examines the hypothesis that the error term, (ε_it − ε_it), is not serially correlated.⁶ Finally, we compute robust two-step standard errors following the methodology suggested by Windmeijer (2005) to correct for small sample biases.

A. Determinants of Inflation Differentials

As discussed widely in the literature, see for instance Honohan and Lane (2003) and Hasan and Alogeel (2008), there are a number of important domestic and external channels that help determine inflationary pressures in the short run and as such are potential determinants of inflation differentials. Figure 1 plots four of the main determinants of inflation: output, credit, non-oil fiscal balance, and NEER growth rates. The sign of the estimated coefficients on output and credit is expected to be positive given their impact on domestic demand. On the other hand, the sign of the non-oil fiscal balance and the NEER coefficients is expected to be negative. As the nominal exchange rate cannot adjust to the movements in the U.S. dollar, we expect that most of the adjustments occur through domestic prices and inflation rates and are captured through the nominal effective exchange rate channel (Table 7). A contractionary fiscal stance through its effect on domestic demand would have a dampening effect on inflation as would an appreciation in the NEER. However, we also explore the effect of government expenditure on inflation differentials which is expected to be opposite to that of the non-oil fiscal balance due to its impact on domestic demand and crowding in of credit to the private sector. There are major differences in the growth of non-oil real output and non-oil fiscal balances as well as credit growth among the GCC countries, with Qatar and the U.A.E. being generally the ones with the highest growth rates. The difference in the evolution of growth rates of these variables appears to follow a similar trend to that of the inflation rate, in which higher growth rates were experienced before the mid-1990’s and after 2003.

Table 7.

GCC Direction of Trade, 1991–2009

(Percent)

Notes: Authors’ calculations based on data from the International Monetary Fund (2010a).

Table 7.

GCC Direction of Trade, 1991–2009

(Percent)

	USA	Europe	Asia	Middle East and North Africa
Bahrain	9	26	22	36
Kuwait	14	34	36	15
Oman	6	22	35	33
Qatar	11	37	30	16
Saudi Arabia	16	35	32	8
United Arab Emirates	9	30	44	8
GCC Average	11	31	33	19

Notes: Authors’ calculations based on data from the International Monetary Fund (2010a).

View Table

Table 7.

GCC Direction of Trade, 1991–2009

(Percent)

	USA	Europe	Asia	Middle East and North Africa
Bahrain	9	26	22	36
Kuwait	14	34	36	15
Oman	6	22	35	33
Qatar	11	37	30	16
Saudi Arabia	16	35	32	8
United Arab Emirates	9	30	44	8
GCC Average	11	31	33	19

Notes: Authors’ calculations based on data from the International Monetary Fund (2010a).

In order to account for the role of structural factors, the share of rent in the CPI basket is included as an explanatory variable. This variable represents the associated inflationary pressures due to supply bottlenecks in the housing market, which have been particularly acute during the recent oil boom. These supply constraints are mainly due to inadequate investments in the past which led to shortages in commercial office space and residential housing. Additional demand pressures arose from opening up the real estate markets to foreign investors in Qatar and the U.A.E. and the inflows of expatriate labor to all GCC countries. A snap-shot of the latest weights used in each country’s CPI basket is reported in Table 8, illustrating the wide differences in the weight assigned to housing.

Table 8.

Share of Food and Rent in the CPI Basket

(Percentage)

Notes: Data provided by country authorities as well as authors and IMF staff calculations.

Table 8.

Share of Food and Rent in the CPI Basket

(Percentage)

	Food, Beverage, and Tobacco	Rent and Utilities	Other
Bahrain	21.0	24.4	54.6
Kuwait	19.0	26.7	54.3
Oman	30.4	21.4	48.2
Qatar	18.1	20.7	61.2
Saudi Arabia	27.7	17.7	54.7
UAE	14.1	39.3	46.5

Notes: Data provided by country authorities as well as authors and IMF staff calculations.

View Table

Table 8.

Share of Food and Rent in the CPI Basket

(Percentage)

	Food, Beverage, and Tobacco	Rent and Utilities	Other
Bahrain	21.0	24.4	54.6
Kuwait	19.0	26.7	54.3
Oman	30.4	21.4	48.2
Qatar	18.1	20.7	61.2
Saudi Arabia	27.7	17.7	54.7
UAE	14.1	39.3	46.5

Notes: Data provided by country authorities as well as authors and IMF staff calculations.

Thus, given the above discussion, the macroeconomic variables that exert short-term influence on the inflation process in country i, z_it in equation (12), is defined as:

where Δneer_it is the growth rate of the nominal effective exchange rate, Δy_it is the growth rate of non-oil GDP, rent_it is the share of rent in the CPI basket, Δcredit_it is credit growth, and Δfiscal_it is the growth rate of non-oil fiscal balances. Economic theory, as discussed above, predicts that non-oil output growth, rent, and credit growth will widen the inflation gap between the countries, while an increase in the growth rates of NEER and non-oil fiscal balances will lead to closing the inflation gap.

A. Analysis Using Monthly Data

We use the proposed system GMM estimator described above to estimate inflation differentials using monthly data:

where $${\mathbf{z}}_{it}^m$$ denotes the variables used in the monthly estimations and is defined as:

where oil_it is oil revenues and the rest of the variables are as defined before. The variables in $${\mathbf{z}}_{it}^m$$ are annualized growth rates. As we have already performed unit root tests on Δx_ijt and confirmed that inflation differentials are in the process of converging, see Section III.A and Table 3, we proceed by estimating equation (17).⁷

Regression [1.2] in Table 9 reports the estimation results using monthly data between January 1995 and June 2010. Oil revenues are a highly significant determinant of inflation differentials, with a ten percentage point increase in oil revenue differentials leading to about a five percent increase in the inflation gap between countries. This effect is indeed expected to explain a large proportion of inflation differentials in the region given the diverse endowments of oil across the six countries. Moreover, this result illustrates the importance of managing the spending of oil revenues over the business cycle to avoid overheating the economy.

Table 9.

GCC Inflation Differentials, 1995M1–2010M6

Source: Authors’ estimations.Notes: The estimation method is two-step system GMM with Windmeijer (2005) small sample robust correction. Standard errors are presented below the corresponding coefficients in brackets. Symbols ^***, ^**, and ^* denote significance at 1%, 5%, and at 10% respectively. The dependent variable is the inflation rate differential between country i and country j; the 15 pairwise combinations are listed in Table 1. All right hand side variables are defined as deviations of country i from country j.

Table 9.

GCC Inflation Differentials, 1995M1–2010M6

Start date:		1995M1	1995M1	2003M1	2003M1
Variables in differentials		[1.1]	[1.2]	[1.3]	[1.4]
Lagged price level, in logs		-0.226***	-0.170**	-0.242**	-0.280**
		(0.047)	(0.062)	(0.112)	(0.098)
NEER growth		-0.447	-0.502	-0.541	-0.361
		(0.385)	(0.376)	(0.353)	(0.362)
Oil revenue growth		0.442**	0.497***	0.625***	0.453**
		(0.156)	(0.162)	(0.185)	(0.170)
Rent weight		0.281**	0.211*	0.504*	0.381*
		(0.116)	(0.103)	(0.251)	(0.180)
Lagged credit growth			0.065*		0.147***
			(0.031)		(0.043)
No. observations		2790	2786	1350	1346
Specification tests (p-values)
(a)Hansen Test		0.877	0.575	0.807	0.763
(b)Serial correlation
	First-order	0.036	0.028	0.102	0.072
	Second-order	0.475	0.200	0.496	0.426

Source: Authors’ estimations.Notes: The estimation method is two-step system GMM with Windmeijer (2005) small sample robust correction. Standard errors are presented below the corresponding coefficients in brackets. Symbols ^***, ^**, and ^* denote significance at 1%, 5%, and at 10% respectively. The dependent variable is the inflation rate differential between country i and country j; the 15 pairwise combinations are listed in Table 1. All right hand side variables are defined as deviations of country i from country j.

View Table

Table 9.

GCC Inflation Differentials, 1995M1–2010M6

Start date:		1995M1	1995M1	2003M1	2003M1
Variables in differentials		[1.1]	[1.2]	[1.3]	[1.4]
Lagged price level, in logs		-0.226***	-0.170**	-0.242**	-0.280**
		(0.047)	(0.062)	(0.112)	(0.098)
NEER growth		-0.447	-0.502	-0.541	-0.361
		(0.385)	(0.376)	(0.353)	(0.362)
Oil revenue growth		0.442**	0.497***	0.625***	0.453**
		(0.156)	(0.162)	(0.185)	(0.170)
Rent weight		0.281**	0.211*	0.504*	0.381*
		(0.116)	(0.103)	(0.251)	(0.180)
Lagged credit growth			0.065*		0.147***
			(0.031)		(0.043)
No. observations		2790	2786	1350	1346
Specification tests (p-values)
(a)Hansen Test		0.877	0.575	0.807	0.763
(b)Serial correlation
	First-order	0.036	0.028	0.102	0.072
	Second-order	0.475	0.200	0.496	0.426

Source: Authors’ estimations.Notes: The estimation method is two-step system GMM with Windmeijer (2005) small sample robust correction. Standard errors are presented below the corresponding coefficients in brackets. Symbols ^***, ^**, and ^* denote significance at 1%, 5%, and at 10% respectively. The dependent variable is the inflation rate differential between country i and country j; the 15 pairwise combinations are listed in Table 1. All right hand side variables are defined as deviations of country i from country j.