THE NEAR EAST AS A GLOBAL COTTON SUPPLIER

Raw cotton has been an important commercial commodity in the Near East for centuries. From the late sixteenth century until the end of the eighteenth century, it was an essential crop for the Ottoman economy. It was used in domestic textile production and shipped in large quantities to Europe, mainly from the Western Anatolian port of Izmir. In Egypt, cotton was used mainly domestically and its cultivation for export began to play an important role in the country's economy only from the early nineteenth century.Footnote ⁸ In general, the expansion of the British and Western European textile industries based on American raw cotton led to a decline in Near Eastern cotton production.Footnote ⁹

Western Anatolia and Egypt's fortunes as cotton suppliers revived during the American Civil War (1861–1865), when raw cotton shipments from American ports to Europe essentially stopped. The crisis, which came to be known as the “cotton famine,” raised concerns about the United States's reliability as a source of supply and resuscitated interest in alternative producers, mainly India, but also the Near East.Footnote ¹⁰ Both Egypt and Western Anatolia saw a considerable extension of the area under cotton cultivation and a spectacular increase in output and exports. Adana, in Southern Anatolia, another significant Ottoman producer of the raw fiber, also gained importance as a cotton supplier during the American Civil War.

At the end of the Civil War, U.S. exports resumed, prices fell, and cotton production in the two regions undertook two separate paths (see Figure 1). Exports of Western Anatolian cotton fell at the end of the century. In contrast, by the 1870s Egypt's output was two and a half times as large as it had been in the previous decade and cotton's importance continued to grow, until it became the country's major export commodity. The average yearly export growth rate was above 18 percent from the 1820s to the early 1880s; for the next three decades, the growth was slower (6.3 percent) but still substantial (Owen 1969, pp. 161, 198–99).

FIGURE 1

WESTERN ANATOLIAN COTTON EXPORTS FROM IZMIR, 1834–1910 (RIGHT SCALE) AND EGYPTIAN COTTON EXPORTS, 1822–1913 (LEFT SCALE) IN TONS

Sources: For Western Anatolia, Quataert (1973); Kurmu? (1987); Mihci and Mihci (2002); and Issawi (1980). For Egypt, Owen (1969).

At the turn of the century, cotton production in Western Anatolia experienced a modest recovery, encouraged by public subsidies for cotton seeds and new infrastructure. The Adana region, where government intervention was coupled with the decisive role of German capital, underwent a more substantial increase in cotton production and exports.Footnote ¹¹ In Southern Anatolia, cotton exports increased over time, in Western Anatolia production edged upwards, but exports rose slowly to an annual average of 30,000 bales in the post-1900 era, a level lower than that reached in the late 1870s (Quataert 1973; Todd 1915). Hence, it appears that an increasing proportion of local cotton was absorbed by the nascent mechanized cotton textile industry.

While the contrast between the rapid growth of Egyptian exports and the more fluctuating path of Western (and Southern) Anatolia informs us about changes in production and trade patterns in the Near East, it is not sufficient to analyze its level of integration with the global economy. As specified in the previous section, I study changes in the behavior of cotton prices over time, in order to gain insights on the process of market integration.

ANALYTICAL FRAMEWORK

The empirical analysis is based on the Takahashi Takayama and George Judge (1971) model of market integration. When transaction costs are zero, the law of one price (LOP) should hold between markets for a homogenous good.Footnote ¹² Since trade costs are almost always present in real world markets, the literature distinguishes between the “strong” and “weak” forms of LOP. The “strong” form of LOP holds in the absence of any trade costs, so that for an identical good a common unique price will exist in spatially separated markets that are linked by trade and arbitrage (Fackler and Goodwin 2001, p. 977). However, owing to the presence of positive trade costs, the LOP in its “strong form” is rarely observed, particularly in the short run. This is because the presence of trade costs produces price differences between markets that cannot be completely arbitraged away, so that price changes in one market are almost never instantaneously transmitted to prices in another market.Footnote ¹³ As a result, economists treat price integration as a long-run relationship, and consider the LOP as an example of a long-run steady-state equilibrium. This has led to the specification of a “weak form” of the LOP, whereby markets are “integrated” when price changes in one market lead to similar proportional changes in price in other markets, when trade costs remain unchanged.

The Takayama and Judge model assumes the traded good is homogeneous. However, cotton is a heterogeneous commodity; thus, the theoretical concept of market integration requires some modification. There are three major subspecies of the cotton plant. Gossypium herbaceum (“Asiatic” cotton) found in India and the Near East has a short fiber with a length ranging from 3/8 to one inch. Gossypium hirsutum (“non-Asiatic” cotton), includes American Upland and has a medium to long-staple fiber with a length varies between one and one and half inch. Gossypium Barbadense (the Peruvian group) includes Sea Island, Egyptian, and Peruvian cottons. This specie had the best kinds of lint and the longest fiber, starting from one and half inch, reaching over two inches for the best qualities.

These varieties are in principle substitutes, but not perfect substitutes. The degree of substitutability among the different typologies of cotton depended on the length, strength, fineness, and character of the lint: while there was a good degree of substitution between American and Egyptian cotton (i.e., between Gossypium hirsutum and barbadense), the substitutability between these two and Gossypium herbaceum was considerably less (Todd 1915, pp. 7–17). In fact, the short-fiber cotton supplied by India and Western Anatolia was only used in coarser cloth, and it was also more difficult to spin and generated a higher share of waste during the production process than the better varieties.Footnote ¹⁴

Because cotton is a differentiated commodity, I cannot expect the strong form of the LOP to hold, even if transaction costs were zero.Footnote ¹⁵ However, the substitutability among the various cotton strains suggests there should be some interrelationship among prices, at least in the long run. I would also expect that an improvement (worsening) in the extent of substitutability would increase (hinder) price convergence. Hence, I test for spatial market integration within the framework of the “weak form” of the LOP.

In my analysis of market integration, I use the price of Western Anatolian and Egyptian raw cotton sold in Izmir and in Alexandria, respectively, and the price of American and Indian raw cotton sold in Liverpool (in the rest of the article, American and Indian cotton implicitly refer to these commodities sold in the Liverpool market). Liverpool was the principal global harbor for the import of raw cotton during the nineteenth century, and the United States the main global supplier, followed by India (Jacks, O'Rourke, and Williamson 2011, see also Figure 2).

FIGURE 2

AVERAGE ANNUAL SHARE IN WORLD COTTON EXPORTS, 1850–1914

Sources: Hanson (1980); Mitchell (1993); Mitchell (2007); Todd (1915); and Mihci and Mihci (2002).

Figure 3 shows the price movements of the four types of cotton from 1845 to 1913. There is a pattern common all the series: a slow rise from 1845 to 1858, followed by a price drop till 1861. The fall in prices was caused by rapid output growth in the United States, which was not met by an equivalent expansion of the textile industry.Footnote ¹⁶ After the huge jump in the early 1860s, coinciding with the American Civil War (1861–1865), prices underwent a downward trend till the end of the century, corresponding to the global depression (1873–1896). The trend reversed after 1896 and price increased until the First World War. Alexandria and Izmir cotton markets clearly matched American and Indian price patterns. There are also times when prices diverge, for example between the mid-1850s and 1860 in the case of Izmir and in the early twentieth century each series goes its own way. In any case, a visual inspection of the data cannot provide an accurate analysis of the degree of interconnectedness (or absence thereof) between Near Eastern and Liverpool prices. Another important characteristic of the series involves price differentials: Egyptian cotton generally fetched more than American owing to its superior quality.Footnote ¹⁷ On the other hand, Western Anatolian, like the Indian fiber, was cheaper than both Egyptian and American cotton because it was of lower quality.Footnote ¹⁸

FIGURE 3

PRICE MOVEMENTS OF AMERICAN, INDIAN, EGYPTIAN AND WESTERN ANATOLIAN COTTON, 1845–1914, IN BRITISH POUND PER METRIC TON

Sources: For Western Anatolian cotton: Consular Reports of the British Foreign Office; Kasaba (1988); Owen (1981); Quataert (1973); Ottoman Agricultural Statistics. For Egyptian: Owen (1969); Richards (1987);and Johnson (1926). For American and Indian cotton: Jacks, O'Rourke and Williamson (2011).

Two crucial factors related to the features of the cotton market need to be examined in order to assess the impact of the dynamics of price transmission: the change in trade policy, in other words the lowering of export tariffs in 1861/62 and the reduction in transport costs. As can be inferred from the analysis presented in the following paragraphs, I expect these two factors to have facilitated the process of market integration, since they generated a decrease in transaction costs.

DATA AND METHODOLOGY

APPLICATION

The first step for understanding the relationship among variables to detect the presence of market integration is to test for the level of integration of the log transformation of each price series, in order to ensure that the regression results are not spurious. Table 2 reports the results using the Jesús Clemente, Antonio Montañés, and Marcelo Reyes approach that allows for two regime shifts.Footnote ²¹ A first set of results allowing the changes in mean to take place at once (Additive Outlier model) show that all prices are non-stationary. If, however, I allow the mean changes to take place gradually (Innovative Outlier model), the Izmir data prove to be stationary (the other series are non-stationary).

TABLE 1

INDEX OF ANNUAL AVERAGE PRICES FOR AMERICAN, INDIAN, EGYPTIAN, AND WESTERN ANATOLIAN COTTON

*1882–1885 for Egyptian prices.

Sources:

Egyptian: index derived from Owen (1969); American and Western Anatolian: Quataert (1973).

According to the unit root test results, the four variables under study show a different level of integration and they all contain structural breaks. To overcome these difficulties, I estimate an autoregressive distributed lag co-integration test (ARDL). This method has been supplemented with the cumulative sum of residuals (CUSUM) and cumulative sum of residuals squared (CUSUMSQ) analysis to detect structural breaks.

The ARDL test is one of the most widespread approaches when variables exhibit different degrees of integration. This methodology also allows the determination of different lag lengths for each variable, unlike other co-integration tests and models, and its bound testing is superior in small samples.

The ARDL co-integration procedure, in other words, tests the absence of a long-run relationship between the dependent variable and the regressors. The validity of the null hypothesis can be examined comparing the F-statistics obtained from the regression with the F-test critical values computed by Kumar Narayan.Footnote ²² The results, reported in Table 3, confirm that the Indian and American cotton prices are co-integrated and that Alexandria and Izmir are co-integrated with both Liverpool series.Footnote ²³ The specifications I used here are with unrestricted intercept and no trend (case III in Narayan 2005).

TABLE 2

CLEMENTE, MONTAÑÉS, AND REYES TEST

* = non-stationary

** = stationary

Notes:

See Figure 3.

Given the existence of co-integration in all markets, I can use the ARDL error correction approach to analyze the long and short-run relationship between Near Eastern and Liverpool cotton prices, following the model suggested by M. Hashem Pesaran, Yongcheol Shin, and Richard Smith (2001). In ARDL models the dependent variable is explained by lags of itself, other explanatory variables, and their lagged values. Including lags creates a dynamic specification of price changes that can allow for lagged adjustment of prices in markets, so that, for example, while prices in two spatially separated markets may not adjust instantaneously to a shock, thus widening price differentials in the short-run, arbitrage will cause this divergence to be temporary and not permanent. The estimation of a long-run model is specified as follows

(1)

(2)

(3)

where P^IND is the price Indian cotton and P^US the price of American cotton sold in Liverpool, while Pⁱ is the price of cotton in Alexandria or Izmir. α₁, α₂, β₁, β₂, γ₁,γ₂ embody the long-run coefficients describing the co-integrated relationship. A dummy variable (war) has been added to equations 1, 2, and 3, in order to take into account the effects of the American Civil War (1860–1865).

TABLE 3

ARDL BOUND TEST FOR CO-INTEGRATION IN LIVERPOOL AND THE NEAR EASTERN COTTON MARKETS

Sources:

See Figure 3.

Following the Schwarz Bayesian Information Criteria, one lag was added to each variable. Once the long-run relationship is established, I can estimate the short-run relationship among domestic and Liverpool cotton prices. This takes the form of an error correction model where the error correction term (ECT) is given by the residuals of the long-run model lagged once and where all variables are expressed in their first difference.

The coefficients of the error correction terms show how fast variables converge to equilibrium and therefore describe the speed of adjustment effect. Hence, by defining how much of the price difference among prices is eliminated in each subsequent period, they represent the short-run speed of adjustment to reach the long-run equilibrium: the closer their value is to −1, the faster is the speed at which convergence takes place. However, since the various cotton strains were not perfect substitutes, I do not expect to observe an ECT equal to −1 (complete price transmission). The presence of an ECT different from zero is a necessary condition for long-run price convergence.

A further step in the analysis of the co-integration relationship consists of testing for the stability of the parameters which define it. This is to ensure that no structural break is present. Unstable parameters would, in fact, undermine the validity of the model and lead to misspecification and biased results.

I follow Pesaran and Shin (1998) and Narayan and Russell Smyth (2005) to find structural breaks in the co-integrated cotton market of Liverpool and of the Near East. The approach applies CUSUM and CUSUMSQ tests to the residuals of the estimated error correction model. The CUSUM test detects systematic changes in the regression coefficients. It utilizes the cumulative sum of recursive residuals based on the first observations and it is updated recursively and plotted against break points. The CUSUMSQ adopts the same procedure, it seeks any sudden departure from the constancy of the regression coefficients. The absence of structural breaks in the co-integration relationship is ensured if the plot of the CUSUM and CUSUMSQ remains within the 5 percent critical bounds.

In Figure 4, I report the CUSUMSQ tests' results for the Liverpool and Near Eastern cotton markets. In the Liverpool market (Indian-American prices), the plot lies outside the 5 percent critical value bounds of the CUSUMSQ test between 1862 and 1898, indicating two structural breaks in the model during those years. These results are of crucial importance since the Liverpool market serves as benchmark for our study of integration.Footnote ²⁴ In the case of the Egyptian-American and Egyptian-Indian relationships, one regime shift was found in 1898 and in 1896, respectively. In the case of Izmir, the CUSUMSQ plot lies outside of the bounds between 1862 and 1896 with respect to the American fiber and between 1867 and 1896 with respect to the Indian one.

The breaks detected between Indian and American cotton in Liverpool signal a change in the relationship between the two prices. Given that such relationship is based on the extent of substitution between the two fibers, the identified regime shifts are likely to reflect a change in the degree of substitutability between the Indian and American varieties.

In the Egyptian-Indian and Egyptian-American tests, breaks were found at the end of the 1890s. This period marked the beginning of a favorable phase for Egyptian cotton production. Productivity was rising, reversing a decade of stagnation due to civil unrest (Schanz 1913; Owen 1969).Footnote ²⁵ The main factor was a rise in demand for Egyptian cotton; in particular, the spread of the process of “mercerizing,” which made the yarn finer and of higher luster so that it resembled silk, revolutionized the industry.Footnote ²⁶ Egyptian cotton, being the longest staple, was one of the most suitable varieties for mercerized threads and fabrics. The rise in demands for Egyptian cotton resulted in stronger linkages with the international market and in a consolidation of trade networks between Egypt and Europe. The end of the 1890s and the beginning of the new century saw an intensification of European involvement in the processing of cotton in Egypt, characterized by a considerable increase in foreign investments in ginning and pressing (Owen 1969).

FIGURE 4

PLOTS OF CUSUMSQ FOR LIVERPOOL, ALEXANDRIA, AND IZMIR MARKETS

Sources: See Figure 3.

In the case of Izmir, three regime shifts occurred. The first phase of integration with the American cotton ended in 1861, while that with Indian prices in 1866: these roughly coincide with the American Civil War. The war led to an unprecedented expansion in cotton production for export on its own. But it was part of a broader trend towards agricultural commercialization in the Near East (Issawi 1966). The American Civil War was particularly important because it was then that cotton became a pivotal export of the Near East. The period after 1896 (the third phase) signals the end of the price depression, after which the empire underwent a period of relative stability. Furthermore, a rise in global demand and the prices for cotton gave new impetus to the Anatolian cotton market (Quataert 1973).

The structural breaks imply that I should dig deeper and estimate long- and short-run parameters before and after the period of the breaks. Doing so allows us to compare the dynamics of the price movements in the various phases and examine the changes in the market relationship between the two Near Eastern regions and the global cotton industry.

Liverpool

The relationship between Indian and American cotton prices in Liverpool was characterized by three phases (see Tables 4 and 5). During the first period (1850–1861), I observe no integration, probably due to the low degree of substitutability between the two fibers. In fact, the vast majority of cotton spun in Lancashire mills was American while Indian short-staple cotton was not considered a good substitute by the British textile industry (Chauduri 1983, pp. 848–49).

The two price series became co-integrated between 1862 and 1898. Integration was spurred on by the American Civil War, when the engineering industry associated with Lancashire mills rapidly adapted mule spinning technology to permit the cotton industry to shift from scarce, longer staple American cotton to the more abundant, shorter staple Indian cotton (Mass and Lazonick 1990, p. 14).

In the last phase (1899–1914), the linkages between the two prices became weaker, reflecting a change in the degree of substitutability between the fibers. The reasons for such developments are not easy to find but one does stand out: the diffusion of ring spinning in Lancashire, especially among vertically integrated firms, at the turn of the century. While ring spinning became economically viable in the mid-1880s, British mills were slow to embrace the new technology and the mule continued to represent the preferred method of spinning (Leunig 2001; Ciliberto 2010). However, during the last period of expansion of the Lancashire cotton industry, after the global price depression, the number of rings increased considerably. In 1902 around 10 percent of mills in the industry adopted the ring, jumping from around 2 percent in 1894 (Ciliberto 2010). Since ring spinning subjected the cotton fiber to more strain, Indian short staple was not suitable for it. Thus, such technological change may have impacted the degree of substitutability between Indian and American cotton, rendering the Indian fiber less valuable.

TABLE 4

ARDL LONG-RUN COEFFICIENTS WITH U.S. PRICES AS EXPLANATORY VARIABLES

* indicates significance at 10 percent.

** indicates significance at 5 percent.

*** indicates significance at 1 percent.

Notes:

Figures in parentheses are t-ratios.

Sources:

See Figure 3.

TABLE 5

ARDL SHORT-RUN COEFFICIENTS WITH U.S. COTTON PRICES AS EXPLANATORY VARIABLES

Notes:

See Table 4.

Sources:

See Figure 3.

Izmir

Within the three phases identified earlier for the Izmir cotton market, the process of market integration varied remarkably over time (see Tables 4–7). In the first phase, the error correction terms of the coefficient of the co-integration equation with the price of American cotton is not significant. The same is true for the price of Indian cotton. Prices were not converging. These outcomes can be attributed to the fact that a large share of raw cotton production was used domestically and its export represented only a residual market. According to a consular report, in 1851 total output was around 30,000 bales, of which 12,000–15,000 were exported and the rest was used in the domestic market. The situation changed drastically in the 1870s, when cotton became, overwhelmingly, an export crop as domestic spinning and weaving activities shrank considerably. Another consular report in the 1870s indicates total production at 60,000 bales, of which more than 51,000 were exported (Issawi 1980). It is not surprising that the raw cotton market responded more to domestic changes, rather than to global ones. Moreover, the lack of a good transport system and high taxes imposed on exports until 1861 (at 12 percent) may have hampered the process of price transmission.

During the second phase (1862–1895 and 1867–1895 for American and Indian cotton, respectively), Izmir was tightly integrated with Liverpool: the improvement was likely due to improvements in transport and infrastructure, particularly the development of the Izmir port and the construction of the Izmir-Aydin railway, which acted as a catalyst in strengthening market linkages. Furthermore, the effect of lowering the export tax (from 12 percent to 1 percent) may have contributed to the process of price convergence. The error correction coefficient of the equation with American (Indian) prices is −0.89 (–0.71), indicating that 89 (71) percent of the disequilibrium in Izmir's domestic cotton prices was corrected each year.

TABLE 6

ARDL LONG-RUN COEFFICIENTS WITH INDIAN COTTON PRICES AS EXPLANATORY VARIABLES

Notes:

See Table 4.

Sources:

See Figure 3.

TABLE 7

ARDL LONG-RUN COEFFICIENTS INDIAN COTTON PRICES AS EXPLANATORY VARIABLES

Notes:

See Table 4.

Sources:

See Figure 3.

In the last phase, from 1896 to 1914, the computed error correction parameter of the equation with American prices is −1.54.Footnote ²⁷ On the other hand, the error correction term of the equation with Indian prices is −0.795. It thus appears that while the process of price transmission weakened with American cotton, it continued with the Indian fiber. Given the good degree of substitutability between the Anatolian and Indian fibers, the slowing down of price transmission with the American series can be connected to the co-integration results in the Liverpool market, and hence to the fact that short-staple cotton may have become a less valuable substitute of the American type.

I conducted the same analysis of international price transmission for Southern Anatolia, using annual prices of Adana's short-staple cotton sold in Istanbul (1876–1908).Footnote ²⁸ The results showed that Adana was integrated with Indian prices during the whole period and with the American series only until 1898.Footnote ²⁹ This outcome provides further strength to our argument on the change in substitutability between short and medium- to long-staple cotton.

CONCLUSION

In this article, I explored the evolution of the process of cotton market integration between Izmir, Alexandria, and Liverpool. The relationship between Near Eastern and international markets was multifaceted, since it encompassed differentiated products, varying export levels, changes in trade and market conditions. Particularly, the fact that cotton is not a homogenous good complicated the analysis of price converge, since it was difficult to identify the effects of changes in substitutability and changes in market integration.

The process of international price transmission in the Near East followed an uneven pattern over time. And the Alexandria and Izmir prices underwent separate paths. Although a large body of literature describes the first wave of globalization as a period in which commodity markets became increasingly more integrated, our study indicates the speed of convergence varied over time.

In Egypt, cotton market integration began in 1845: while strengthening in relation with the American fiber, the process of convergence with Indian prices weakened at the turn of the century. In Western Anatolia, I observed almost the opposite scenario. Market integration started later, with the American Civil War, and it continued intensifying only with respect to Indian cotton. On the other hand, convergence with American prices slowed down considerably.

Improved market integration between the Near Eastern and the Liverpool markets between the 1860s (1845 for Egypt) and mid to late 1890s was undoubtedly aided by lowering transaction costs, particularly connected with transportation and communication improvements worldwide. However, at the turn of the century, some of the established price linkages changed: integration worsened between Anatolian and American and between Egyptian and Indian cotton.

The reasons for such deterioration in price convergence between some cotton types are not easy to extract. I know that there was neither an increase in transport costs nor changes in market conditions or trade policy. I also know that Anatolian cotton was a close substitute of Indian and Egyptian of American throughout the period under analysis. Therefore, some of these divergences in integration are likely to be attributed to variations in the extent of substitutability between short- and medium- to long-staple fibers. Thus, technology changes in spinning that impacted the degree of substitutability between Indian and American cotton in our benchmark market in Liverpool had repercussions on the integration of the Near Eastern markets.