China’s International Trade

China implemented a set of highly restrictive international trade policies and remained relatively closed in the early nineteenth century. Trade with European countries was small compared to other economic activities and faced strict regulations. In 1840, conflicts between British merchants and local Chinese officials led to the outbreak of the Opium War. After losing the war, the Chinese Government in 1842 was forced to sign the Treaty of Nanking, which reduced tariffs and opened five ports to international trade. Later, additional treaties were signed, which established more ports, a British-governed Maritime Customs Service, and a flat low tariff rate. By the beginning of the twentieth century, 45 ports were opened (Yan Reference Yan2012). The real trade value stayed roughly the same from 1868 to 1900 and steadily grew afterward. In 1930, the real trade value was 7.72 times the value in 1900.Footnote ³

Cotton yarn was among the first group of foreign commodities accepted by the Chinese. As early as 1880, “many parts of the Chinese Weavers prefer Foreign Cotton Yarn to the Chinese [ones]” (China Maritime Customs 1881, p. 24). Its import quantity increased steadily from 1880 to 1913, with an average annual growth rate of more than 9 percent.Footnote ⁴ It is estimated that approximately 64 percent of the cotton yarn consumed in China before WWI was imported (Fang Reference Fang2011, p. 287).

Chinese Cotton Yarn Industry

Handicraft textile production was an important secondary activity for farmers across China by the nineteenth century. Indian Cotton (or Asian Cotton) was first introduced to South China before 500 BC and spread to North China in the thirteenth century. To produce a piece of cloth, a worker first cleaned the cotton to remove cotton seeds and dirt, and carded it for spinning. The second step was to make cotton yarn from the prepared cotton using spindles, and the third step was to weave cloth from cotton yarn using looms. The traditional handicraft sector required heavy labor inputs. To produce a piece of cloth with a width of 0.4 meters, length of 6.67 meters, and weight of 1 kilogram, a very skilled Chinese worker spent approximately six working days, with one, four, and one days for the three steps, respectively (Xu Reference Xu1992, p. 51). Spinning was so costly partially because most handicraft spinning wheels only had a single spindle. Therefore, when textile machines allowed for multiple spindles at the same time, manufactured yarns easily demonstrated advantages over the handicraft ones. In fact, Yan (Reference Yan2011) believes that productivity in the machinery yarn sector was 80 times that in the handicraft spinning sector when measured by output per labor, whereas productivity was only four times higher when using machinery looms to replace handicraft ones.

Facing shocks from imported manufactured yarn, the Chinese started to build their own manufactured textile industry during the second half of the nineteenth century. Early efforts came from the government, including setting up manufactured textile companies in Shanghai and Hubei (Shanghai Jiqi Zhibuju and Hubei Zhibuju) in the late 1880s. These factories faced difficulties in funding and management (Perkins Reference Perkins1967) or survived at the expense of prohibiting the entry of private firms (Institute of Shanghai Economy Studies 1958a); they had limited impact on the overall economy. In 1895, as the Treaty of Shimonoseki allowed foreigners to erect factories in China’s ports, the resulting total number of foreign firms increased from 603 in 1895 to 1,006 in 1900 and 2,863 in 1911 (China Maritime Customs 1913). These firms brought foreign capital, technology, and management skills that stimulated the growth of the domestic industrial sector. The government also encouraged the entry of private enterprises during this time.Footnote ⁵

As Figure 1 already suggests, the boom of the manufactured textile industry took place after 1913. From 1913 to 1930, Chinese yarn dramatically increased from 0.53 million bales to 2.4 million bales (Kubo Reference Kubo2004). Because most products were consumed by the local market, the ratio of yarn produced in China to domestic yarn consumption increased from 36 percent in 1913 to 102 percent in 1930 (Fang Reference Fang2011, p. 287), indicating that China switched from being a yarn importer to a self-sufficient producer and even an exporter.

Along with the growth of total production, the location of textile firms also changed over time. Figure 2 depicts the distribution of ports and firms in 1907 and 1925. The triangles indicate the location of treaty ports, which remained the same during the study period.Footnote ⁶ The circles represent the textile firms, among which the darker ones suggest more firms. At the beginning of the period under study, textile firms were mostly located close to Shanghai. Ultimately, they spread to multiple other cities, especially the ones located close to ports. This is consistent with the hypothesis that shocks from international trade might have affected firm entry.

Figure 2

DISTRIBUTION OF TEXTILE FIRMS AND PORTS IN 1907 AND 1925

Sources: The number of textile firms is from Yan (2011). The name of ports is recorded by China Maritime Customs (1907). The base map is from the China Historical Geographic Information System, version 6 (2016).

Local characteristics were likely to drive firms’ locations as well. Textile firms in China heavily relied on intermediaries in finance and trade to acquire inputs, especially machinery and raw cotton. A survey in 1918 on textile firms in Jiangsu Province, where most firms located, showed that all textile machinery was imported, of which 90 percent was from Britain (Fang Reference Fang2011, p. 97). Another major input for textile production was raw cotton, whose quality directly related to the quality of yarn. While some firms deliberately chose to locate in counties suitable for raw cotton growth, most raw cotton was not consumed locally in as early as the 1910s. Rather, the harvested raw cotton was transferred and sold to large cities by cotton brokers (hua hang) through well-functioning commercial networks. As a typical example, 80 percent of the raw cotton grown in a small county in North China was carried and transferred to large cities 100 to 150 miles away, leaving only 20 percent for local consumption.Footnote ⁷ Some firms also imported raw cotton from India or the United States. These transactions usually required financial institutions, which formed an exchange network of bank notes and provided credit.

Shortage of capital was another obstacle for the Chinese firms, one that haunted every stage in the growth of a textile firm. Entrepreneurs usually faced difficulties in fully collecting the required capital when starting a business. Once profits were earned, shareholders often requested dividends immediately rather than leaving sufficient reserves for future investments. For example, from 1902 to 1914, the textile firm Dasheng gave 64.7 percent of its profits to shareholders and 13.5 percent to staff, while only 16.8 percent was saved for the firm’s development (Gu Reference Gu2015, pp. 45–46). This firm had to rely on loans for activities involving large monetary transactions, including purchasing raw cotton and acquiring machinery. This was, in fact, a common practice for most textile firms, suggesting that intermediaries in finance were essential complements for firms’ operations.Footnote ⁸ Figure 3 plots the location of textile firms in 1925 and banks in 1912. Dark areas indicate that the county had at least one textile firm and larger circles represent more banks. Counties with more banks before the war tended to have more textile firms at the end of the study period.

Figure 3

DISTRIBUTION OF TEXTILE FIRMS AND BANKS

Sources: The number of textile firms is from Yan (2011); banks were domestic banks with western institutions (in other words, domestic modern banks) from the Nongshang tongji biao (Reports of Statistics on Agriculture and Commerce). The base map is from the China Historical Geographic Information System, version 6 (2016).

In addition to access to inputs, political chaos and foreign concessions might have affected the location of firms. The examined period witnessed constant political turmoil that might have discouraged private investments. During this time, foreign concessions were enclaves managed by foreigners and under western institutions. They usually had better protection of property rights and attracted more industrial investments.Footnote ⁹ For example, the largest foreign concession, Shanghai, also contributed 41 percent of national manufacturing output in the 1930s and owned more than 50 percent of cotton spindles during the study period (Ma Reference Ma2006, Reference Ma2008).

Measures of the Textile Industry

Limited information is available for the cotton textile industry in the 1910s because most surveys of this industry were conducted after 1919 (see Zeitz (Reference Zeitz2013) for a review). I collect records of firm entry and exit from a secondary source and aggregate the data to the county level. To check how sensitive the results are to data quality, I compare estimates using separately constructed records.

The baseline analysis relies on Yan’s (Reference Yan2011) collection of entry, acquisition, merger, and exit of each mill. These data record the founding time, years of actual operation, changes in names, and year when each mill ceased operating. I aggregate the data to the county level and obtain the number of textile mills in operation each year. Because this dataset traces the entry and survival of individual textile firms, it captures the “extensive margin.” There were 99 firms during the examined period. As an alternative measure of the textile industry, I also use the investments of new textile firms compiled by Du (Reference Du1991) to ensure that the analyses are not sensitive to the choice of data source or outcome variable. Du (Reference Du1991) independently collected information of industrial firm entry before 1927 from official reports, Japanese surveys, and firms’ archives. This dataset reports the name, year of establishment, location, initial capital invested, and founders of 109 textile firms.Footnote ¹⁰ Because it only observes firm entry and no exit, I use it as a robustness check.

In addition to using historical data, I also examine two individual firms’ experience to provide qualitative evidence about the entrepreneurs’ constraints and decision processes. These firms are chosen due to their rich archives and distinct background. Table 1 summarizes the owners, year established, and scale of the two firms. The first firm, Dasheng, was among the earliest domestic textile firms in China. Its founder, Zhang Jian, had a government-related background. This firm performed well before 1922 but experienced serious crises when the market was down from 1923. I cite Gu’s (Reference Gu2015) detailed analysis of Dasheng’s economic archives. The second firm, Shenxin, was founded by two merchants, the Rong brothers, in Shanghai in 1916. It entered the market later but became the largest Chinese textile enterprise before 1937. Its archives, including account books, balance sheets, and entrepreneurs’ diaries, are compiled and published by the Shanghai Social Science Academy (1980). I also supplement the qualitative analysis with descriptive studies of textile firms’ daily operations (Fang Reference Fang2011), financial constraints (Yan Reference Yan2011), performance during the war (Tokihiko Reference Tokihiko2010), and commercial network (Brasó Broggi Reference Brasó Broggi2016).

Table 1

INFORMATION OF INDIVIDUAL FIRMS

Sources: Yan (Reference Yan2011); Shanghai Social Science Academy (1980).

The Trade Shock

I capture the trade shock using information on two dimensions. The time-series variation reflects fluctuations in the import of cotton yarn. I also exploit spatial variation, coming from differences in access to international trade at the county level.

The overall volume and quantity of traded cotton yarn are obtained from the annual reports of the China Maritime Customs Service (CMC), which was a government agency operated by British officers during the examined period. Its main task was to collect tax revenues from China’s international trade. The CMC published yearly reports on the quantity and value of each traded commodity, tax revenues, and trade value of each port. Every ten years, it published another series of reports, including a summary of economic conditions in each port and the average trade values of important commodities, including cotton products. These reports provide detailed information about international trade in pre-modern China, and they are acknowledged by scholars to have high-quality data.Footnote ¹¹

I collect data from the CMC to capture different dimensions of textile trade. For the world price, I digitize the import quantity and value of cotton products at the national and the port level and use the information to calculate the value per unit of yarn as a proxy for the yarn price. The analysis uses data at the national level, but as Online Appendix Figure 1 illustrates, the per unit value of imported yarn moved very closely across different ports.Footnote ¹² I also collect the number of foreign vessels entering China to illustrate the impact of WWI on trading activities.

To measure access to international trade, I construct a transportation network for China by combining information on major transport routes, including rivers, railroads in 1913, and historical courier roads. The network of river and courier routes comes from the latest version of the China Historical Geographic Information System (CHGIS) (Harvard Yenching Library 2016).Footnote ¹³ The railroad network is digitized using the map by the United States Central Intelligence Agency (1954) and assigned the order of availability based on accomplishment times recorded in Zhongguo tiedao jianshe shilue (1876–1949) (A Brief History of Railroad Construction in China (1876–1949)) (Zhang Reference Zhang1997).Footnote ¹⁴ After combining all the routes and setting a transportation speed to each type of route, I calculate the transportation cost between a given county to all ports on the map and locate the port with the lowest cost in terms of time. This measure is similar to the “effective distance” used in Juhász (Reference Juhász2018). Figure 4 depicts the cost to ports in days. More than 60 percent of the counties could access ports within three days.

Figure 4

TRANSPORTATION COSTS TO PORTS (IN DAYS)

Sources: The road network is composed of four types of transportation mode: railroads, rivers, historical postal routes as a proxy for major roads, and access roads that are direct lines to connect counties not accessible using the first three transportation modes.The assigned speeds per day of railroads, rivers, major roads, and access roads were 1,200 km, 150 km, 75 km, and 40 km, separately. Online Appendix A provides a more detailed discussion about the choice of the speed.

To address the concern that access to large ports might have had different effects than access to small ports, I also calculate the cost to six major treaty ports, selected based on their importance in cotton yarn trade. Located across China, these six ports together contributed more than 40 percent of yarn imports during this period. In fact, the import of cotton yarn was not dominated by an exceptionally large port but rather evenly distributed across at least 15 ports.Footnote ¹⁵ The dark triangles in Figure 4 represent the location of major ports.

Local Conditions

In addition to access to trade, local conditions were likely to affect firm entry. I collect data to quantify factors that entrepreneurs often considered when setting up a firm during this period. These factors are financial institutions, geographic suitability for textile production, and political chaos.

The first set of variables measures financial institutions. In this period, financial institutions were composed of traditional banks, Chinese banks structured under western institutions (hereafter modern Chinese banks), and foreign banks. Online Appendix B presents a brief discussion of their differences. The data on the first two types of banks come from the survey in 1912 by the Bureau of Agriculture and Commerce (1914), which reports capital for modern Chinese banks at the branch level and the number of traditional banks in each county. I use the data in 1912 because this is the only year where the branch information is available. The location of foreign banks’ headquarters is collected from Waiguo zaihua gongshang qiye cidian (The Universal Dictionary of Foreign Business in Modern China) (Huang Reference Huang1995). A benefit of using the cross-sectional distribution of financial institutions is to address the concern that the location of banks might have responded to firm entry. I check the correlation of bank capitals from 1912 to 1921 to ensure that banks’ locations in 1912 are a valid proxy for the later locations.Footnote ¹⁶

I control for whether a county’s indigenous characteristics were suitable for textile production using two indicators. The first indicator is whether a county was a historical center of handicraft textiles. I use the location of handicraft textile centers in the eighteenth century presented by Guo (Reference Guo1990) and identify approximately 50 counties as historical textile centers. The second indicator is geographic suitability for cotton growth. I collect this information from the Global Agro-Ecological Zones (GAEZ), a data portal developed by the Food and Agriculture Organization (FAO) (2012).Footnote ¹⁷ The original data report suitability in eight levels for a 50 × 50 km² cell, and I calculate the mean of all cells in a given county. The data show that northern counties are more suitable for raw cotton growing (Online Appendix Figure 2).Footnote ¹⁸

The last set of control variables addresses political turmoil. The county-level locations of political chaos are constructed from Zhonghua minguo junshishi (The Military History of the Republic of China) (Jiang Reference Jiang2009), which describes the time, location, parties involved, and leaders of army battles in major domestic military conflicts since 1911.Footnote ¹⁹ The locations of foreign concessions, the peace shelters during chaos, are from Yan (Reference Yan2012), who provides a list of all foreign concessions in China.

Table 2 reports the summary statistics for the variables. The first panel presents the variables capturing industrial performance at the county level. The second panel lists the measure of the trade shock. The last two panels show the control variables, including county characteristics and indicators for political chaos and foreign concessions.Footnote ²⁰

Table 2

SUMMARY STATISTICS

Sources: Sources are provided in the table.

Explaining Patterns in Firm Entry

Rawski (Reference Rawski1989) and Reynolds (Reference Reynolds1975) observe that there were lower capital formation and slower domestic textile industry growth during WWI compared with the period after the war. The econometric analysis of this paper supports this statement and finds that firm entry was greater after WWI than during the war. What was the reason for this delayed boom?

I first address the conjecture that the pattern was driven by foreign investments. China was open to foreign investments during the examined period and there was a surge in the number of Japanese textile firms after 1921 (Tokihiko Reference Tokihiko2010). It is possible that the later entry of firms was driven by foreign investments that paused during the war and flooded in after the war was over. To consider the entry of foreign firms, I divide the sample by firm nationality and re-run the regression. The first three columns of Table 5 report the estimations using a full sample of counties for the Chinese firms and the ones after dropping Shanghai. Because most foreign firms clustered in Shanghai, I only report the full sample results for the Japanese and the British firms in Columns 4–7. The postwar coefficient estimates are statistically significant for the Japanese firms and not for the British firms, but in both cases, the absolute values are much smaller than the ones for the Chinese firms, suggesting that foreign investments were not the driving forces for late entry.

Table 5

IMPACT OF WWI AND ACCESS TO TRADE ON THE NUMBER OF TEXTILE FIRMS IN CHINA, BY NATIONALITY

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors, reported in parentheses, are clustered at the provincial level. The regression covers the period from 1907 to 1925. The dependent variables are the numbers of textile firms in county i at time t by nationality (plus one and raised to the natural log form). WWI and PostWWI are dummy variables. τ is constructed travel time in days (plus one and raised to natural log form) from a county to its closest port.

Source: Author’s calculations from Yan (Reference Yan2011) and CHGIS (2016).

In fact, the reason for the postwar rather than in-war entry was likely due to the postponed arrival of machinery. China was unable to produce textile machinery and relied heavily on imports from overseas, especially from the United Kingdom (Fang Reference Fang2011, pp. 96–100). Firms’ archives show that, while the shipping of machines from Europe to China already took substantial time, the war further stalled machinery arrival. In 1917, when Shenxin intended to expand production by purchasing new machines, difficulties caused by the war hindered the business. Rong Desheng, the owner of Shenxin, wrote in his diary, “In 1918 … [we] intended to increase spindles, but we could not find a company to purchase from … Mu Ouchu (the owner of another textile firm) obtained machines with 25,000 spindles from the US, but their arrival was also delayed. [The planned] ShenXin II ordered 10,000 spindles from the Great Britain, and these spindles came after the war. In preparation for the arrival of new machines, we uninstalled the old machines and waited, which led to considerable lost opportunities” (Shanghai Social Science Academy 1980, p. 59). Similarly, Zhang Jian ordered machines from the British firms for a new factory, Dasheng III, in 1915, but the machines did not arrive until 1921 (Gu Reference Gu2015, p. 175).

The trade statistics suggest that many firms might have faced similar difficulties. Figure 8 presents the total number of orders with the British manufacturers and the import values of textile machinery in each year. The dashed line shows the total number of orders with the British manufacturers, constructed based on Wright’s (Reference Wright2011) records of textile machinery transactions. These data recorded individual transactions involving major textile machinery producers in Britain, with codes for major supplies, trading countries, firm identifiers, number of frames and spindles, and order and shipping times (Saxonhouse and Wright Reference Saxonhouse and Wright2010). I select the ones with China. Because Britain was the main producer of textile machinery in China during this time, these orders serve as a good proxy for the general demand for textile machinery from firms in China. The total number of orders surged in 1909, 1914, and 1920.Footnote ³³ The dotted line depicts the import values of textile machinery. Starting from a low level, the import values doubled from 1918 to 1919, still less than 2,000 Haikwan taels, and surged to more than 30,000 Haikwan taels in 1922. While the number of orders mostly parallels the trend of the machinery importation, there were relatively more orders from 1914 to 1919, compared to the actual machinery shipped.

Figure 8

THE IMPORT VALUE AND ORDERS OF TEXTILE MACHINERY.

Sources: The import value is from Hsiao (1974, p. 47, Table 2, “Textile Machinery”). The British orders are from Wright (2011).

In Wright’s (Reference Wright2011) transaction-level data, a subsample of transactions with Platt, a major supplier of textile machinery to the Chinese firms, also records the order time and the shipping time of each order.Footnote ³⁴ I calculate the gap between the two times for these transactions. The lag was very long when China just started its textile industry in 1889. It fell to less than one year according to the only order containing this information from 1900 to 1914. However, the several transactions with this information from 1914 to 1919 on average were delayed by more than three years, except for one order that was shipped within a year. The average lag was around two years for transactions from 1918 to 1920 and gradually dropped to around one year after 1921.

I also collect firm-level capital and the actual inputs of the Chinese textile firms for the years 1916 to 1921.Footnote ³⁵ The capital remained the same from 1916 to 1917 and increased steadily from 1917 to 1920, with an average annual rate of approximately 21 percent. In contrast, the number of spindles remained the same until 1920. Online Appendix Figure 3 plots capital invested and spindles at the firm level each year. This suggests that, although firms increased investments during the war, the textile machinery did not change until the war was over.

Suggestive Evidence on the Role of Finance

The previous section established the relationship between the war-inducing trade shock and the entry of textile firms. This section provides suggestive evidence of the additional conditions for firm entry, including access to finance and trading companies, geographic suitability for raw cotton, a history of handicraft textile production, and political conflicts. I especially emphasize the importance of access to financing, which facilitated trade and provided capital for the textile firms.

I control for the prewar level of these factors and conduct a triple-difference analysis:

$${y_{it}} = {\beta _0} + {\beta _1}WW{I_t} \times {\tau _i} + {\beta _2}WW{I_t} \times Cha{r_i} + {\beta _3}WW{I_t} \times {\tau _i} \times Cha{r_i} \\+ {\beta _4}PostWW{I_t} \times {\tau _i} + {\beta _5}PostWW{I_t} \times Cha{r_i} + {\beta _6}PostWW{I_t} \times {\tau _i} \\\times Cha{r_i} + {\mu _i} + {\sigma _t} + {\rm{ }}{\varepsilon _{it}}$$

where the war dummies, τ_i, µ_i, and σ_t, have the previous interpretations. Char_i is a vector that includes a set of local characteristics that did not vary with time. The interactions of the war dummies and Char_i measure how local characteristics were correlated with heterogeneous effects of the trade shock on port counties during and after the war. The triple-interaction terms WWI_t × τ_i × Char_i and PostWWI_t × τ_i × Char_i capture how the heterogeneous effects of associated local characteristics varied with trade cost. The coefficients of the triple-interaction terms, β₃ and β₆, indicate whether the association with local factors differed when firms had different access to international trade.

The local characteristics include financial institutions in 1912, historical textile centers in the eighteenth century, a proxy for suitability for cotton production, and measures for political stability. Because the financial sectors and the industrial sectors were often coincident, the regression provides more suggestive evidence rather than a causal interpretation. I check the pre-trend to show that regions with different access to banks had no clear differences in the growth paths. Online Appendix Figure 4 shows that, after controlling for national time trends and county fixed effects, counties with and without banks had similar pre-trends in terms of firm entry.

Financial institutions might have promoted firm entry through facilitating trade or providing capital. To shed light on the mechanisms, I also collect the location of trading companies that served as brokers during international commodity transactions. If financial institutions promoted firm entry through trade solely, the coefficient should shrink dramatically after controlling for these trading companies.

Table 6 reports the regression results. Column 1 shows the estimates without controlling for the trading companies. The interaction terms between the bank variable and the war dummies are positive and statistically significant, suggesting that, when counties were located at ports, those with more financial institutions had more firm entry than those without financial institutions. The coefficient estimates show that a 1 percent increase in bank capital at the county level was associated with 0.00467 percent more entry of firms during the war and 0.0226 percent after the war. The triple-interaction terms are negative, showing that higher transportation cost reduced the positive effect on firm entry due to access to financial institutions. Specifically, a 1 percent increase in cost to ports reduced the previous positive effect by 0.00397 percent during the war and 0.0186 percent after the war. The result suggests that the positive effect from exposure to international trade and access to finance would decrease to zero after the logged value of trade cost is greater than approximately 1.2, which is about 3.3 days.

Table 6

IMPACT OF WWI AND LOCAL CHARACTERISTICS ON THE NUMBER OF TEXTILE FIRMS

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors, reported in parentheses, are clustered at the provincial level. The regression covers the period from 1907 to 1925. The dependent variables are the numbers of textile firms plus one in natural log form in county i at time t. WWI and PostWWI are dummy variables. τ is constructed travel days from a county to its closest port (plus one and raised to the natural log form). The variable bank is the bank capital in 1912 plus one and raised to the natural log form. I also control for the distribution of foreign banks, traditional Chinese banks, concessions, suitability for cotton production, and warfare that occurred in a county in a given year.

Source: Author’s calculations from Bureau of Agriculture and Commerce (1914), CHGIS (2016), Huang (Reference Huang1995), Yan (Reference Yan2011, Reference Yan2012), Guo (Reference Guo1990), and Jiang (Reference Jiang2009).

Columns 2–4 report the results after controlling for trading companies. The interaction terms of trading companies and the war dummies are positive, suggesting that these companies were positively associated with firm entry. As expected, the coefficient estimates of interaction terms between the bank variable and the war dummies decrease, but they mostly remain statistically significant. This implies that at least part of the banks’ function was probably to provide capital. The results are robust if I replace the war dummies with per unit value of cotton yarn, as reported in Online Appendix Table 2. I also use longitudinal data from 1912 to 1920 instead of the cross-sectional data in 1912 to measure bank distribution in Online Appendix Table 3, and the results are similar.

Firms entering the market after the war had missed the golden years and suffered from the market downturn in the 1920s. In 1919, the market price for yarn was even higher than during the war years, and firms continued to enjoy the profitable markets. Starting in 1922, however, the price of yarn fell and raw cotton became expensive. During this market downturn, debts and interest became increasingly burdensome. According to Dasheng’s account book, the share of interest in its total costs and fees was nearly 10 percent during the first several years, and it soon surged to more than 20 percent. In 1920, 30 percent of the firm’s expenditures were interest payments, which led to a major crisis when the firm failed to pay its loans in 1924 (Gu Reference Gu2015, p. 35). Dasheng’s experience was not unique. In fact, the pressure of debts forced many entrepreneurs to sell their firms to large enterprises, banks, or other merchants. Among the 99 firms established before 1925, in Yan’s (Reference Yan2011) record, at least 54 of them changed ownership, with 21 taken over by banks, 10 by Japanese merchants, 2 by British merchants, and the rest sold or rented to other owners (Yan Reference Yan2011, Appendix 1). Firms’ experience suggest that, while the trade shock increased the total number of spindles in China, more than half of the firms failed to survive. Most small firms finally became part of the large enterprises.