Testing the Productivity Bias Hypothesis in Middle East Countries Testing the Productivity Bias Hypothesis in Middle East Countries

Divergence of the purchasing power parity from the equilibrium exchange rate is attributed to various factors. Productivity differentials between the countries are said to be one of the main sources, which lead to productivity bias hypothesis. The hypothesis suggests that a relatively more productive country should experience a real appreciation of its currency. This research aims at testing the hypothesis in Middle East countries using the time series data over the period of 1970-2015 and by employing ARDL approach to cointegration. The econometric results support the hypothesis is only in the case of Bahrain, Kuwait and Saudi Arabia. This research also provides policy recommendations on the basis of empirical results.


Introduction
The Purchasing Power Parity (PPP) is the oldest theory of exchange rate determination which asserts that exchange rate between currencies of two countries is equal to the ratio of the general price level of the said countries. The validity of the PPP has been tested empirically many times and the results are not conclusive, as discussed in Bahmani-Oskooee and Nasir (2005). Divergence of the PPP from the equilibrium exchange rate is attributed to various factors. Productivity differentials between the countries are said to be one of the main sources, which lead to productivity bias hypothesis (PBH).
The PBH, which is also known as Balassa-Samuelson Hypothesis (BSH), simply suggests the appreciation of a currency in a relatively productive country. According to this hypothesis, a country with high productivity growth also experiences high wage growth, which causes a rise in prices and consequently, the real exchange rate appreciates. The origin of the hypothesis lies in the seminal works of Balassa (1964) and Samuelson (1964), where both authors independently observed that differentials of productivity growth lead to real exchange rate appreciations. Balassa (1964) advocated that due to the higher level of productivity in production of tradable goods relative to non-tradable goods (services), the exchange rate will be overvalued in terms of PPP in countries with relatively high production of tradable goods. Prices of non-tradable goods or services will be greater in countries with a higher level of productivity because of the relatively low level of productivity in the service sector. The larger gap between relative productivity gives rise to a greater gap between prices and as a consequence, deviations take place between PPPs leading to currency overvaluation. It is also noted that a more productive country is supposed to have higher standards of living, which causes higher prices of consumed goods and services. As a result, the increasing gap between prices leads to an appreciation of the more productive country's currency. Samuelson (1964) also argued that productivity differentials were a main contributing factor of overvaluation of the US dollar in the 1960s. Since then, many studies have been conducted in an attempt to test the hypothesis.
This study aims to fill empirically the gap in the literature for the Middle East countries by employing time series data and ARDL approach to cointegration. As far as this study is concerned, the PBH has not been exclusively tested in the case of the Middle East countries. This paper is motivated further that the underlining cause of the real appreciation should be understood in depth so that the exchange rate policy can be designed accordingly. Section 2 presents a brief review of the literature.
Section 3 outlines the econometric methodology. Section 4 reports the econometric results. Section 5 concludes. Officer (1976) and Bahmani-Oskooee and Nasir (2005) Balassa (1964), who was the first to empirically test the hypothesis, found significant results by comparing ratios of PPPs to exchange rates based on 12 countries from the Organization for Economic Co-operation and Development (OECD). However, De Vries (1968) rejected the hypothesis using the sample of 64 countries. Using the data of 12 OECD and 19 Latin American countries, the study of Clauge and Tanzi (1972) did not find support for the hypothesis. Officer (1976) presented a new specification using the productivity differentials on a sample of 15 industrial countries and revealed significant results for the hypothesis. Clauge (1988) used the sector-specific model to evaluate the hypothesis for 19 Latin American countries and the results were in favour of it. Falvey and Gemmell (1991) extended the sector -specific models to general equilibrium framework and presented an empirical evidence for the hypothesis. The study of Bahmani-Oskoee and Niraoomand (1996), which adopted the same model as Officer (1976), failed to support the hypothesis. Bahmani-Oskooee and Nasir (2001) tested the hypothesis by pooling cross-sectional data from 68 countries over the 1960-1990 period and the results supported it. Hsieh (1982), using the OLS (Ordinary Least Squares), displayed the first time series evidence for the hypothesis. The studies of Rogoff (1992) and Bahmani-Oskooee (1992), which both employed Engle-Granger (1987) cointegration approach, are particularly notable considering the time series properties of the variables and providing significant results for the hypothesis. Strauss (1995) employed the Johansen and Juselius multivariate cointegration approach for 14 OECD countries and lent support for the hypothesis. Using ARDL approach to cointegration to a data set of 44 countries, Bahmani-Oskooee and Nasir (2004)  Ali (2015), Cardi and Restout (2015) and Wang et al. (2016).

Panel Studies
The first panel study on the PBH was conducted by Asea and Mendoza (1994), which used the data set of 14 OECD countries over the period of 1970-1985. The research of Asea and Mendoza (1994) used different categories of dependent and independent variables and revealed significant results for the hypothesis. The panel study of Chinn (2000), which consisted of 9 Asian-Pacific countries, supported the hypothesis. Egert et al. (2003)  It is evident that empirical results may vary with the econometric techniques, data quality, model specification and data span.

Model and Econometric Methodology
This study adopts the model of Officer (1976), hence we form the following long-run relationship between real exchange rates and productivity differentials, in double logarithmic linear form as: where RERt is real exchange rates, expressed as in which is the price level in country i(US). EX is the equilibrium exchange rates defined as number of i's currency per unit of dollar. PRODt refers to the productivity differentials defined as . Thus, the productivity of country in i is and is the productivity in US. ε is the classical error term. t stands for time period. Equation (1) postulates that if a more productive country is to experience a real appreciation of its currency in the long-run, it is expected that the slope parameter, should be positive.
The long-run relation in Eq.
(1) should incorporate the short-run dynamic adjustment process in order to provide insights of adjustments between time periods. To this extent, Engle-Granger (1987) cointegration approach can be utilized in the first instance. Then, Eq. (1) becomes as follows: where represents change, γ is the speed of adjustment parameter and − ε is the one period lagged error correction term, which is estimated from the residuals of The ARDL cointegration approach involves two steps for estimating the long run relationship. The bounds testing procedure is based on a Wald type (F-statistics) which is also the first step of the ARDL cointegration method. Accordingly, a joint significance test that implies no cointegration under the null hypothesis, (H0: Lastly, if the F-statistic falls between these two sets of critical values, the result is inconclusive. W-testing procedures in the Pesaran et al. (2001) approach are considered to be pre-testing for cointegration. Moreover, this stage of testing is very sensitive to lag selection criterion and lag lengths. As a consequence, it is quite likely that the establishment of a cointegration relationship may fail due to wrong selection criterion or selected lag length. To overcome this possible shortcoming, we follow Kremers et al. (1992) and Banerjee et al. (1998) who proved that a negative and significant ECt-1 could be used as an alternative evidence of cointegration if the Engle-Granger (1987) approach fails to establish a cointegration relationship among the variables. Therefore, this study will also utilize the results from the error correction model to establish the existence of cointegration should the pre-testing stage of Pesaran et al. (2001) fail to do so.

Empirical Results
Eq. (3)   However, in the latter case, the slope estimates of Egypt, UAE and Yemen appear to be negative, indicating that the PBH does not hold for them. In the case of Kuwait, it is seen that the estimated slope coefficient is statistically significant and is greater than zero which validates the PBH.
[Insert Table 1 About Here] Regarding the results of the ARDL approach to cointegration, the long-run slope estimates of Bahrain and Saudi Arabia are positive and statistically significant, suggesting the existence of the PBH. However, the slope estimates of Oman and Qatar are statistically insignificant, indicating that we cannot draw any statistical inferences on these countries. Hence, we eliminate these countries from further analysis. These results demonstrate that there is only partial support of the hypothesis in the Middle East countries since only 3 out of 17 estimates display statistically significant and positive values of the long-run slope parameters. Within these three countries, the impact of the PHB is the strongest in the case of Bahrain, showing that 1% rise in the relative productivity leads to 0.73% appreciation in real exchange rates. Table 2 presents the order of ARDL procedures, value of lagged error correction terms, some standard regression diagnostics such as autocorrelation, functional form, heteroscedasticity, normality and summary results of the overall residual stability tests.
It appears from the results in Table 2 that the diagnostic tests of Bahrain, Kuwait, and Saudi Arabia are also statistically satisfactory which support the reliability of the econometric results. Regarding the lagged error correction terms with significant PBH in Table 2, Saudi Arabia has the highest lagged error-correction term of -0.24 which suggests that any disequilibrium between the currencies of USA and Saudi Arabia will be eliminated within around four years.
[Insert Table 2 About Here]  Brown et al. (1975). S indicates stability and NS indicates instability.

Concluding Remarks
The PPP hypothesis holds providing that none of the assumptions behind it is being violated. However, in the literature, the productivity differentials between the countries have been identified as one major factor for the deviations of the equilibrium exchange rates which gave rise to the PBH. The PBH suggests that there exists a positive association between exchange rates and productivity differentials implying that higher productivity causes a real appreciation of a country's currency.
This research tested the validity of the PBH for 17 Middle East countries using the ARDL approach to cointegration. The econometric results from this research reveal that the PBH is being validated in the case of only 3 Middle East countries, namely Bahrain, Kuwait and Saudi Arabia.
We suspect the failure of the PBH in the rest of Middle East countries may be attributed to some macroeconomics factors such as the impact of globalization and free trade movements that many developing countries have been experiencing, in addition to the consequences of various government policies in the form trade, exchange rate and development, which have not been included in our study.
As far as the policy recommendations are concerned, it is crystal clear that the countries should develop economic policies that would lead to a rise in productivity, especially in the sectors of tradable and non-tradable goods in order to gain international competitive advantage in real exchange rates in the long-run. Those policies should be very comprehensive and sustainable so that the gains from them would last for a long time. Macroeconomic policies aiming at increasing productivity may range from different simple tax incentives to sophisticated education of labour force. To this end, for example, the quality of labour in the sector of tradable goods plays a crucial role for raising international competitiveness. Improving the labour quality with education will also increase the productivity of this production factor.
Similarly, research and development expenditures may be utilized specifically to increase the productivity of production factors of capital and technology.