The feasibility of currency union in Gulf Cooperation Council countries: A business cycle synchronisation view

1 Introduction

Regional economic integration in the Gulf region has improved in the last two decades, and we expect greater convergence in macroeconomic policy. This study empirically investigates the feasibility of creating a currency union among Gulf Cooperation Council (GCC) countries (i.e., Bahrain, Kuwait, Oman, Qatar, Saudi Arabia and the United Arab Emirates [UAE]), following closer monetary cooperation in recent years. One of the objectives of these countries is the establishment of a common currency the khaleeji. The political and economic will to create this single currency for all GCC countries is a subject of great interest. The similarities among these economics in terms of both economic structure and their attachment to the US dollar provide an additional argument for this proposed union. In addition, GCC countries have since December 2000 adopted an exchange-rate regime pegged to the US dollar—with the exception of Kuwait, which since May 2007 has repegged to an undisclosed weighted basket of international currencies. Indeed, these countries “have successfully maintained pegs among themselves for two decades despite their failure to meet a number of important OCA [optimum currency area] criteria” (Willett, Al-Barwani, & El Hag, 2010, p. 1716). Furthermore, these countries’ macroeconomic similarities have allowed for considerable harmonisation in terms of inflation rates (see Figure A5 in Appendix B).

One of the main determinants of the success of a monetary union is whether the common monetary policy set by a common central bank is suitable for all member countries. This suitability in turn depends on the degree to which business cycles are synchronised across the countries involved. “According to the classical OCA criteria, two countries or regions would benefit from forming a monetary union if they are characterised by high similarity of business cycles” (Babetskii, 2005, p. 106). For this reason, we need to determine the degree of business cycle synchronisation among GCC countries.

In a monetary union, an independent counter-cyclical monetary policy cannot be tailored to stabilise individual cyclical fluctuations. Since the seminal study by Mundell (1961), economics has identified different criteria by which to assess the welfare effects of a monetary union; factors mobility, wage and price flexibility, and similarity preferences among member countries need to be considered. Business cycle synchronisation among member countries is one of the pre-conditions of a monetary union: indeed, the lower the level of business cycle synchronisation among the countries looking to form a monetary union, the higher the costs of this union will be. The endogenous approach with an OCA (Frankel & Rose, 1998) considers that a monetary union affects the nature of domestic business cycles.1 The establishment of a currency union increases trade intensity, and this in turn leads to an increase in output correlation—and thus to an increase in business cycle synchronisation. The new economic geography stresses that trade integration may increase country-level specialisation, leading to lower business cycle synchronisation. Empirical studies dedicated to industrialised countries show that trade intensity increases business cycle comovement among them (Clark & Van Wincoop, 2001). For emerging and developing countries, research results are mixed (Calderon, Chong, & Stein, 2007). Differences in the patterns of specialisation and bilateral trade are the main explanations for these results. While trade among industrialised countries is largely of an intra-industry nature, in developing countries, interindustry trade tends to prevail. Consequently, the relationship between trade intensity and the alignment of business cycles is positive in the former case, but negative in the latter.

The introduction of a single GCC currency began with the establishment of a Monetary Union Committee in 2002; the underlying concept was born much earlier, however, with the GCC Charter and the Economic Agreement in 1981. Low regional intracountry trade constrained the related countries to postpone pursuance of this single currency until after the development of certain integration stages, including the Customs Union (launched in 2003) and the common market (2005). However, GCC countries rely heavily on expatriate labour. A labour market that features a Kafil (citizen guarantee) system creates a concentrations of foreigners in the private sector and citizens in the public sector. Labour immobility is compensated with a high foreign labour percentage of total employment (i.e., 55% in Bahrain and greater than 80% in Kuwait, UAE and Qatar). We can conclude that the labour market criteria were satisfied (Willett et al., 2010). The integration process seems to be very slow, but heading in the right direction. At the end of 2005, the GCC approved the convergence criteria, to ensure financial and monetary stability. The monetary criteria include inflation rates and sufficiency of foreign cash reserves, while the financial criteria include the ratio of the annual government finance deficit to gross national product (GNP), as well as the ratio of public debt to GNP.

The objective of this study was to assess the feasibility and desirability of creating a GCC currency union, based on a framework of OCA theory. Since Mundell (1961) first introduced the concept of the OCA, a vast body of literature has developed around it. Focusing on the relationships among potential OCA members, they examine the extent of trade, mobility of labour and capital, the risk-sharing system and similarities in terms of shocks and cycles. Without denying the importance of the first three criteria, we focus on the last one: it is generally accepted that comovement among business cycles is a crucial criterion in a country's decision to join a currency union.

We focus not only on the feasibility of a monetary union, but on which circumstances country-pairs best comove, and in which period. Relying on a new measure of synchronisation, we identify various subperiods in which comovement of business cycles change in the six GCC economies. We are interested in two types of cycles—namely, short and long term. These correspond to a 40-quarter cycle length (i.e., the Juglar fixed investment cycle of 7–11 years) and a 10-quarter cycle length (i.e., the Kitchin inventory cycle of 3–5 years), respectively.

A higher degree of business cycle comovement in the long run is a necessary condition of monetary union; on its own, however, it is not sufficient. A common monetary policy must include a short-term effect. The primary cost of adopting a common currency is the loss of monetary independence, but this loss is minimised as cycles synchronise across countries. This study is novel in three respects: (i) it examines the dynamics of economic integration among GCC countries; (ii) it differentiates between the short and long terms; and (iii) it provides for us the level of feasibility of each GCC country joining a monetary union.

The remainder of this paper is organised as follows. Section 2 presents stylised facts about GCC economic integration and the related literature. Section 3 explains the methodology adopted in this study. Section 4 analyses the main empirical results. Section 5 concludes.

2 Stylised Facts about GCC Economic Integration, and the Related literature

Even after three decades, the GCC has not yet reached its primary goal of creating a single market with a single currency. Compared to countries in the European Union (EU) or in East Asia (EA), the GCC group of countries is characterised by a low level of intra-GCC trade. However, this fact will not be decisive in the degree of business cycle synchronisation given their highly similar economic structure (i.e., they all rely heavily on the oil sector). Additionally, GCC governments intervene to regulate internal prices when the oil market is in recession.

Taking the OCA argument of Mundell (1961) as our starting point, we analyse the synchronisation characteristics of GCC countries, to determine the desirability and feasibility of establishing a currency union there. As stressed by Mundell (1961), the suitability of a common regional monetary policy requires that all countries share the same state (i.e., expansion vs. recession). This fact implies strong business cycle synchronisation—hence our interest in these OCA criteria.

It is possible that integration increases the degree of industrial specialisation, which could in turn lead to greater inconsistency in terms of business cycles. The effect of specialisation can be reduced if bilateral trade is essentially intraindustrialist rather than interindustrial. Frankel and Rose (1998) and Rose (2002) stress that economic integration stimulates the correlation of business cycles. This higher correlation rests on the intensification of bilateral trade among a specific group of countries. Fidrmuc (2004) argues that increased bilateral trade (i.e., intra-industrial) intensifies the degree of synchronisation of business cycles.

Compared to that on other regional integration experiences, the literature on the GCC monetary union is not abundant. In addition, the theoretical and empirical literature is not conclusive about business cycle synchronisation among GCC countries. One strand of literature uses a different empirical model to test the desirability of such monetary integration. Zaidi (1990) assesses the desirability of such a monetary union by examining the capacity to hold a common attitude towards inflation and unemployment, stating that “the requirement of [a] long-run balance-of-payments equilibrium under fixed exchange rates imposes constraints on domestic macroeconomic policies” (Zaidi, 1990, p. 764). The author shows that in response to unanticipated money growth, both the variance of real output and unanticipated money vary across the GCC economies; greater coordination is needed, if these economies are to achieve similar patterns of variability in monetary growth. Laabas and Limam (2002), whose work is based on G-PPP theory, argue that GCC countries are not yet able to satisfy all necessary preconditions for establishing a currency union. Basher and Fachin (2014) studied the demand for broad money in the GCC economies, and they conclude that there is a low level of desirability for a monetary union given the low-level homogeneity (i.e., poolability) in the monetary demand function. Based on a general equilibrium model, Mohamed and Irandoust (2014) suggest that the best strategy for GCC countries is to peg their common currency to special drawing rights (i.e., a US dollar/euro basket) to the US dollar.

The second strand of the literature proposes a test of monetary union desirability by examining business cycle synchronisation. The predominant empirical methodology in this strand is to use the structural vector autoregression (SVAR) model to analyse the effects of macroeconomic shocks, or the symmetry of shocks, to test monetary union feasibility (Abu-Qarn & Abu-Bader, 2008; Bayoumi & Eichengreen, 1994; Kandil & Trabelsi, 2012; Louis, Balli, & Osman, 2008). In particular, Eichler and Karmann (2011) focus on the feasibility of monetary union in five regions, and find that GCC countries have converged over time, in spite of the low level of symmetry of shocks. Interestingly, Rafiq (2011) quantifies, using a structural factor model, the existence of synchronous change among GCC countries. In fact, three countries—namely, Bahrain, Oman, and the UAE—share a common factors shocks for output fluctuations.

Kandil and Trabelsi (2012) conclude that GCC countries are still far from forming a viable monetary union; however, they find a subgroup—comprising Saudi Arabia, the UAE and Qatar—where correlations of underlying disturbances could help foster a monetary union. By undertaking an SVAR exercise, Jean Louis, Osman, and Balli (2010) test the feasibility of the US dollar as an anchor for the GCC currency. They find common inflationary responses to US monetary policy, but divergent responses for non-oil outputs. For this reason, they suggest that pegging the GCC currency to a basket that includes the euro and the US dollar might be an optimal choice.

Cevik (2014) identifies the degree of non-hydrocarbon business cycle synchronisation across GCC countries. Using an asymmetric band-pass filter and a mean-corrected concordance index, he finds a low heterogeneous level of business cycle synchronisation during the 1990–2010 period. That study's non-dynamic empirical methodology seriously affect its results, which were negatively affected (i.e., via skewed data) by the first Gulf war (i.e., the Kuwait invasion) in the early 1990s. Meanwhile, Djennas, Benbouziane, and Djennas (2013) evaluated the relative influence of each variable used, while taking into account extreme bound analysis. In addition, they used a wavelet technique to identify coherence between pairs, and to remove noise from business cycle index values. In their study, the overall period is decoupled exogenously into three subsample periods (i.e., 1980–89, 1990–99, 2000–11), to determine variations in business cycle index; we find the same subperiods in the study of Benbouziane, Djennas, and Cherif (2013). Djennas et al. (2013) stress not only that GCC countries diverge in terms of business cycles, but that divergence has attenuated since the global economic crisis. Willett et al. (2010) calculate an annual lagged correlation (based on one and two-year lags) and observe a low magnitude of divergence; moreover, the coefficients of a number of certain correlational pairs are negative. As the gross domestic product (GDP) of the GCC countries rely heavily on oil and gas export, those GDP values will fluctuate as the oil price changes; for this reason, any lagged correlation of GDP or related data will be weak. Using 10-year rolling window calculations of three year-to-year correlations between Saudi Arabia and each of five small countries, Willett et al. (2010) found low levels of business cycle correlation.

Benbouziane and Benamar (2010) tested the possibility of an OCA for the six GCC countries. They used a multivariate threshold autoregressive model to test the non-linear relationships that could exist among the studied economies. Furthermore, they observed the existence of two subgroups. The first subgroup comprised the UAE, Oman and Bahrain, while the second group comprised Saudi Arabia, Qatar and Kuwait. Marzovilla (2014) found two subgroups, divided on account of their economic structure. On one hand, there are countries that adopt a diversification strategy (i.e., the UAE, Bahrain, Qatar and Oman); on other hand, there are countries that strongly embrace the mining and oil sectors (i.e., Saudi Arabia and Kuwait). The two specialisation models adopted by the two groups seem to reduce intragroup economic cycle differences, but increase intergroup differences. The author suggests that the strategy of pegging to the US dollar seems to adversely affect GCC economics, given the divergence between the cycle of the GCC countries and that of the United States. For this reason, the use of a regime that pegs to a basket of currencies could help reduce pass-through. Therefore, these results must be interpreted with caution. First, the author takes the overall period as being homogenous, or exogenously chooses subperiods; however, the region has been known to sustain multiple crises and shocks. Second, there is a link of causality between diversification strategy and reduced business cycle synchronisation. Third, among the six countries exhibiting a high level of correlation in economic cycles, we find three countries that belong to different groups.

Benbouziane et al. (2013) analysed the relationship between business cycle synchronisation and some related variables (i.e., bilateral trade, trade openness, trade specialisation, economic specialisation, deposit interest rate differentials, official exchange-rate fluctuations, fiscal deficit differential, financial openness, monetary policy, current account balance, gross national savings as a percentage of GDP and oil exports). They identified the main determinants of business cycle synchronisation using extreme bound analysis. Their results show that the best candidates are Saudi Arabia, Qatar and the UAE. In addition, they compared business cycle synchronisation within country-pairs, in terms of evolution in time and degree of synchronisation; they did so using the wavelet method. With the second technique, an absence of coherence and convergence was found among the group of countries studied. Based on continuous wavelet analysis, Aloui, Hkiri, and Nguyen (2016) argue that real growth rates co-move in the majority of GCC economies in the short term (i.e., fewer than 5 years). Saudi Arabia and the UAE are also synchronised for the long term.

Our methodology outperforms those seen in the literature, in two main respects. First, by using our dynamic model, we can detect variability in the integration process. Second, we can divide variability in the synchronisation business cycle into two components (i.e., short and long term).

3 Empirical Methodology

Essentially, the feasibility of a monetary union depends on whether the common monetary policy will be suitable for all members, in the sense that all countries will need the same monetary policy (i.e., expansionist vs. restrictive). To achieve concordance, we check for business cycle synchronisation among all GCC country-pairs. If the business cycle in the short and the long term is synchronous across countries, this means that achieving a common monetary policy is feasible, and the desirability of a monetary union is high.

3.2 Evolutionary spectral analysis: definition

The evolutionary aspect of the spectrum relates to the non-stationarity in series, which follows an oscillatory process (Priestley, 1965, 1988, 1996):

(1)

where for each w the sequence {A_X(w, t)}, as it is according to t, admits a maximum Fourier transform (in module) in zero with {Z_X(w)}, an orthogonal process on[−π, π], , , and μ_X(w) a measurement. We conclude that the evolutionary spectral density of {X_t} is the function S_X(w, t), defined as follows:2

(2)

where . The variance of {X_t} at t is:

(3)

Let we suppose a two zero-mean real stochastic process (X_t; Y_t). Let S_X(w) and S_Y(w), , be the spectral density functions and S_XY(w) be the cospectrum. Each of the two series can be written as:

(4)

where:

(5)

where the overline indicates complex conjugation. As is well known, we have:

(6)

Equation (6) verifies this inequality coherence, .

is the coherence in the neighbourhood of w:

(7)

is an analogue to the coefficient of the correlation between two samples and is interpreted as it.

3.2.1 Evolutionary cross-spectra (S_XY(w, t))

Priestley (1988) extends the theory of evolutionary spectra to the case of bivariate non-stationary processes. Consider now two oscillatory component processes, (X_t; Y_t). We can write:

(8)

where:

(9)

Let and denote, respectively, the families of the oscillatory and . We define the evolutionary power cross-spectrum at t with respect to the families and , dH_XY(w, t), by:

(10)

According to Priestley (1988), in the non-stationary case, the cross-spectrum is time-varying and is defined as dH_XY(w, t). By virtue of the Cauchy–Schwarz inequality, we have:

(11)

We can now write:

(12)

where S_XY(w, t) is the evolutionary cross-spectrum density function.

(13)

where:

(14)

are the real time-varying cospectrum (i.e., the gain) and the time-varying quadrature spectrum (i.e., the phase), respectively. ℜ and ℑ are the real and imaginary parts of the time-varying cross-spectrum.

3.2.2 Estimation of the coherence of the non-stationary spectra

First, consider the observable bivariate time series (X_t; Y_t), which are not necessarily stationary. Their time-varying spectra can be defined via S_X(w, t) and S_Y(w, t), respectively. Therefore, the time-varying cross-spectrum given by Equation (13) has a polar representation:

(15)

These statistics allow us to calculate the time-varying cross-amplitude in the following manner:

(16)

The time-varying magnitude squared coherence:

(17)

3.3 Time-varying coherence test

Some techniques have been developed to test multiple structural breaks. We adopted the tests of Bai and Perron (1998, 2003a, 2003b) to detect a mean shift in TVCF. Using GAUSS software, we obtained estimates by running the codes of Bai and Perron. The selection of this type of model was motivated by the TVCF characteristics. Indeed, analysis of the graphical patterns of these statistics (see Appendix A) indicated that it is affected only by breaks in the mean. Using Bai and Perron (1998) allowed us to endogenously determine break-point dates, when the changes in coherence between series were significant. We are interested in the break point in the coherence between two economies. The break point is defined as the change in the underlying relationship of the two economies, and it occurs in response to an exogenous event or a change in economic policy.

3.3.1 The model and estimators

We considered the following mean-shift model with m breaks, (T₁, …, T_m):

(18)

for , and , where T is the sample size. is the TVCF in the neighbourhood of the w_j frequency. μ_i are the means, and u_t is the disturbance at time t. The break points (T₁, …, T_m) are explicitly treated as unknown. Based on the ordinary least squares (OLS) principle Bai and Perron (1998), we estimated the vector of the regressor's coefficients μ_j by minimising the sum of the squared residuals . Let denote the resulting estimate. Substituting it into the objective function and denoting the resulting sum of squared residuals as , the estimated break-point dates are such that:

(19)

where the minimisation is taken over all partitions such that , where O takes some value.3

3.3.2 Test statistics and the model selection criteria

This test locates multiple breaks without imposing any prior expectations on the data. The procedure estimates unknown regression coefficients together with the break points when T quarters are available. To determine the number of break points, we use the Bayesian information criterion (i.e., BIC) as suggested by Yao (1988) and defined as follows:

(20)

where is the number of unknown parameters. The author demonstrates that, for a normal sequence of random variables with shifts in the mean, the number of breaks can be consistently estimated.

Bai and Perron (1998) introduced some asymptotic critical values for the arbitrary small positive number and the maximum possible number of breaks (M): , , , and . For our empirical computation, we choose and we used the algorithm of Bai and Perron (1998, 2003b) to obtain global minimizers of the squared residuals.

We estimated time-varying coherence to explore the dynamic properties of the cyclical component of the convergence process. These statistics will provide a measure of persistence of comovement in the region for each country-pair. We also investigated whether an international economic crisis and/or regional political tension has had any impact on the degree of business cycle convergence, by endogenously testing break-point dates through the use of the Bai and Perron test.

4 Empirical Results

5 Conclusion

In this paper, we proposed a new measure of business cycle synchronisation, to assess the feasibility of a Gulf Cooperation Council (GCC) monetary union. Our main finding is that not all GCC countries have a synchronised business cycle in the long term. However, for the short term, all GCC country-pairs show a high level of synchronisation in the most recent subperiods. On the one hand, the United Arab Emirates’ (UAE's) new role in the regional trade platform has allowed it to strengthen the long-term business cycle comovement. On the other hand, other GCC country-pairs not involving the UAE have shown a decline in the two most recent subperiods. With respect to the other criteria; however, a GCC monetary union does not appear to be feasible. Therefore, a high level of short-term business cycle synchronisation does appear to offer the promise of regional integration. Overall, given that the countries within the region do seem to demonstrate a positive synchronisation trend, taken together with their integrates labour markets and common economic structures and cultures, establishing a common currency does appear to be within reach.

However, these results must be interpreted with caution, for two main reasons. On one hand, as suggested by the recent world economic crisis, business cycle synchronisation in the GCC seems still to be driven by external shocks, rather than by an autonomous regional process. Even if the GCC's trade linkages today with the Group of three countries (Japan, USA and China) are less important than they previously have been, the GCC remains vulnerable to international economic fluctuations, as seen with the recent fall in oil prices. On the other hand, countries may have different initial responses to shocks and react symmetrically to shocks with one or two period lags. For this reason, further business cycle synchronisation studies are needed, which test the reaction of each country to different shocks; such research would help affirm the feasibility of a monetary union in the region. In addition, the region must sustain the trend of positive synchronisation by introducing additional reforms—especially with regard to monetary authority, as a central bank must be independent, if it is to assist in establishing strong regional institutions.