STRUCTURAL TIME SERIES ANALYSIS OF U.S. COTTON EXPORTS

This study employs a structural time series method to model and estimate U.S. cotton exports. The results show that the fluc- tuations observed in U.S. cotton exports have transitory (cycle), secular (trend), and seasonal characteristics. The estimated structural relationships after accounting for the impact of the unobserved components indicate that U.S. cotton exports vol- ume responds positively to higher international price relative to domestic price of cotton and negatively to real exchange rate volatility. The study also confirms that the U.S. government export-marketing loan program has a significant and direct effect on U.S. cotton exports. A difficult competitive environment has led the U.S. textile industry to shift from the production of high-valued end-use products and to focus more on production and exports of low-valued products such as yarns and fabrics (Hudson and Eth- ridge, 2000). This new strategy has contributed to the vitality of textile industries in the developing world, especially in Asia, the Caribbean basin, and Mexico leading to increased demand for cotton in these areas. The volume of shipment was evalu- ated at 11.9 million bales for the 2002-2003 marketing year, which represents 38.7% of the world cotton exports (USDA, 2003). Most of the exports are directed to countries where the U.S. has a sizable share of the cotton markets with Mexico (96.3%), China (60.7%), Turkey (72.3%), Pakistan (56.0%), India (32.3%), Canada (100%), South Korea (31.7%), Thailand (29.3%), Indonesia (37.6%), and Taiwan (44.4%). As textile industries in these countries are increasingly searching for high quality cotton fibers and reliable supply sources, the U.S. shares are expected to remain high in the foreseeable future. While cotton exports volume is generally determined by a favorable parity between international price and domestic price of cotton, i.e. exports flow is expected to increase when international price of cotton is above domestic price, economic fluctua- tions such as downturns in foreign textile manufacturing activities and currency realignments remain determining factors. For instance, a relative appreciation of the dollar with respect to the currencies of the U.S. textile trading partners, 13% since the 90's according to MacDonald (2002), and the Asian financial crisis have been identified as the main reason for the decline in U.S. cotton exports between the 1996/97 and 2000/01 marketing years. Though it is widely accepted that robust industrial activities have positive effects on exports, empirical evidence about the effects of exchange rate and exchange rate volatility on trade flows, including commodity exports, is not unequivocally estab- lished. Studies conducted in this area have led to contradictory conclusions despite hypothesizing that exchange rate volatil- ity hinders exports (Bini-Smaghi, 1991), Chowdhury (1993), and Arize (1995, 1996), among others. The conflicting empirical results from previous studies are due to specification problems. For instance, while volatility is per- ceived as a risk, in many studies, it is derived as a moving average of the growth rate of real exchange rate. The moving aver- age process only accounts for the expected volatility and thereby does not fully measure risk (Bini-Smaghi, 1991). Chowd- hury (1993) also pointed out other procedural flaws in earlier studies stemming from a failure to account for the possible integration and long run relationship between exports volume and most of its determinants, including exchange rate and world price. When two variables integrated of order one establish a long-run relationship, either an error correction model (ECM) or first-order autoregressive distributed lag model (ADL(1)) can be used to assess the short and long run dynamics between the variables. However, these methods provide little to no information with respect to the trend, cyclical, and sea- sonal components of the series. To circumvent the procedural limitations as previously outlined, this study proposes a structural time series approach that uses the Kalman filtering procedure in a state-space form to model the components (trend, seasonal, and cycle) and the struc- tural relationships between exports and its determinants. Modeling the unobserved components along with the economic variables affecting exports volume provides the possibility to isolate the different sources of fluctuations and thereby better assess the contribution of each set of variables to exports flow.