Classical Worship

Masters were worshipped in China from as early as the Western Zhou Dynasty (1046–771 BCE), when schools offered sacrifices to the deceased masters to foster students’ respect for the learning of the older generations. From the Han Dynasty (202 BCE–220 CE) onwards, the worship of masters became gradually institutionalized. Emperors conferred the title of “sage” on eminent masters and enshrined them in the Imperial Temple of Culture—the highest honor for scholars in imperial China. Their works were promoted as classics and employed as the principal curriculum in schools and civil examinations. This allowed the imperial authorities to control the direction of scholarship and the empire’s ideology. From the Tang Dynasty (618–907) on, sage worship became a regular activity in imperial schools. Official sacrificial ceremonies were usually held twice a year, following rigorous rituals enacted by the imperial court (Wilson 2002).

The private worship of classical masters was more prevalent than regular official sacrifices. This worship was mainly conducted in local temples dedicated to specific sages. The temples were typically built in county seats or academies and were open for scholars and students to visit and worship the sages whose statues were placed in the middle of the main hall. Sophisticated sacrificial ceremonies also took place in these temples (see Figure A1 in Online Appendix 2 for illustrations). To some extent, classical worship became the “religion” of the Chinese learned class. Historical narratives indicate that many literati sincerely practiced this worship. They believed in the superiority and sanctity of the sages’ learning, and that their devotion could “connect” them to the sages’ spiritual world (Lu Reference Lu2016).

The classical education (and examination) system further consolidated classical worship. The imperial authorities used examinations to select qualified officials from the sixth century until their abolition in 1905. Given that candidates required a mastery of classical learning to pass the civil service exams, such learning dominated school education and academia. Sons of literate families were trained to recite the texts of the Four Books and the Five Classics from the age of five or six years old. Scholars’ academic standing was based on their mastery and exegesis of classical texts. Considerable talent and resources were therefore directed toward classical studies. Since the Han Dynasty, classical education has been highly valued in educated families, as described in the proverb “bequeathing your sons a piece of classic book is better than leaving them a basket of gold” (Books of Han Dynasty). In the seventeenth to eighteenth centuries, scholars of the Kaoju (evidential studies) school focused on “rediscovering” the antiquarian texts to reform the metaphysical neo-Confucian scholarship (Elman Reference Elman1984). While there were some “cultural entrepreneurs” in traditional China, their innovations were largely in the tradition of the classical canons; they were much less revolutionary and enlightening than their European counterparts, such as Francis Bacon, Isaac Newton, and Martin Luther (Mokyr 2016). Even the most renowned Qing mathematician, Mei Wending (1633–1721), asserted that “the accumulation of human knowledge is merely a token of the ancients’ superior merit” (Jami Reference Jami2012, p. 220).

In sum, continuous state promotion made classical worship so persistent in historical China. The classical scholastic system functioned so well that Chinese elites had few incentives to develop a new or modern one, which would be costly, uncertain, and likely to jeopardize their vested interest in the classical system (Qian Reference Qian1985; Mokyr 2016). Therefore, classical worship in historical China seems to be more institutionalized and overwhelming than that of Britain at the time, where many intellectuals exhibited distinct classical tastes but still appreciated new learning in the market of ideas.

Modernization and Its Classical Fetter

Finally, China’s classical beliefs encountered an unprecedented challenge from the industrial civilization after its forced opening up in 1842. The Western influence stimulated some open-minded Chinese elites to introduce new thoughts and industrial technologies, providing China with the impetus for industrialization. From the late 1850s, new industrial firms experienced steady growth in the textile, machinery, chemical, and various other industrial sectors (see Figure A2 in Online Appendix 2).

However, classical beliefs and mindsets still drove many conservative elites to seek ways of self-strengthening from their intellectual ancestors and to resist the tide of modernization. They believed Western technologies were superior only at the “implementation” level but were not comparable to Chinese learning in “spirit.” As the Qing scholar-official Wo Ren (1804–1871) submitted, Chinese wisdom in the self-cultivation of propriety could defeat Western weapons and machines. Likewise, Zhang Shengzao opposed the establishment of Tong Wen Guan, the first modern institute for teaching Western languages and sciences, on the grounds that the new curriculum would direct students’ interest from the essentials (classical learning) to the trifles (technologies) (Zhongguo Shixue Hui 1961). Such cultural triumphalism made them what Landes (Reference Landes2006) called a singularly “bad learner.”

Even modern intellectuals were more or less overshadowed by classicism. For example, pioneers of the Westernization (Self-Strengthening) Movement from the 1860s to the 1890s still upheld the principle of zhongti xiyong (Western learning for use under the body of Chinese learning). This sentiment echoes those of their Kaoju predecessors, who absorbed the Jesuit sciences into the Chinese classics rather than developing Chinese sciences using European methods (Ma Reference Ma2021). Likewise, the leader of the Wuxu Constitutional Reform, Kang Youwei (1858–1927), sought legitimacy for his reform proposal from Confucius.

The defense of the classical camp did not cease after the imperial regime collapsed in 1911. The classicists initiated movements to protect Chinese orthodoxy in response to the New Culture Movement that propagated the sciences and modernity in the 1910s to 1920s. These quarrels resembled the battles between the Ancients and the Moderns in seventeenth- and eighteenth-century Europe. In the next section, I quantitatively assess whether the classical ethos systematically inhibited China’s industrialization.

Classical Worship

To measure the strength of classical worship in each prefecture, I use the number of temples that were built to worship the sages—acknowledged masters of classical learning. Scholars were selected as sages via a process of nomination and approval from other renowned scholars of the time (Chu Reference Chu2008); the imperial authorities then conferred their status. Therefore, the distribution of sages across prefectures is unlikely to have been influenced by the Emperor in the same way that the Catholic Church strategically selected more saints in regions that had more Protestants (Barro and McCleary Reference Barro and McCleary2016).

I use the temple measurement for two reasons. First, temples constitute the tangible presence of classical worship in historical China. Local scholars and elites built them to carry out sacrificial visits, ceremonies, and other memorial services. More temples were erected in regions where the admiration of classical wisdom prevailed. Second, since the temples were the primary sites conducting classical worship, they would further consolidate the ethos of classicism in a region. Historical narratives indicate that local scholars viewed the temples as the “nodes” that linked them to the ancient masters by transmitting the masters’ spirit and wisdom (Yang Reference Yang1997).

I collected data on the sage temples from the Unified Chorography of the Qing Dynasty (Jiaqing Chongxiu Daqing Yitong Zhi). Compiled by the Qing court in 1820, the Chorography records all the temples in each prefecture, including each temple’s location and which figures it features (see Figure A3 in Online Appendix 2 for an illustration). I manually checked the biography of each figure and identified 336 temples that worshipped sages.Footnote ² The Chinese imperial authorities conferred a total of 181 sages between circa 25 and 1911. I exclude 41 sages from the sample because they either were not scholars or were conferred after 1820. See Shandong Tongzhi (Provincial Gazetteer of Shandong), compiled in 1934, for biographical records of all the sages.

The sage temples were unevenly distributed across China proper (Figure 1); about 57 percent of prefectures had at least one, up to a maximum of 11 (see Figure A4(a) in Online Appendix 2 for the histogram of the number of sage temples). More temples were built in regions that were economically and educationally more developed. The sage temples were built between the sixth century BCE and 1820, peaking from the twelfth to the seventeenth centuries. Therefore, the number of temples in 1820 reflects the accumulation of classical worship from the distant past and is thus less subject to the reverse causality of industrialization after the mid-nineteenth century. However, this measure may not capture later changes in the strength of classical worship, such as being undermined by modern forces in some regions. In the absence of systematic data on the distribution of sage temples after 1820, I assume that classical worship accumulated from the millennium-long history persisted until the sample period (1858–1927).

Figure 1 DISTRIBUTION OF SAGE TEMPLES (1820) AND INDUSTRIAL FIRMS (1858–1927)

Notes: The strength of classical worship is reflected in the number of sage temples that were built to worship the acknowledged masters of classical learning by 1820. Industrialization is shown as the number of Chinese domestic industrial firms established between 1858 and 1927. The map includes 269 sample prefectures of China proper based on the Qing Dynasty administration in 1820 (CHGIS, version 6).

Source: See the text.

Another concern is that the temples recorded in the Chorography may not include all temples that ever existed, as some may have disappeared by 1820. Yet the survival of the temples is also informative and may strengthen the measure of classical worship, since temples are more likely to persist in stronger worship regions. That said, the temple measure does not exclude measurement errors. For instance, sages in some prefectures received forms of worship other than temples (e.g., in scholars’ homes, writings, and hearts), which this measure cannot capture.

For comparison, I also counted the distribution of other temples built to worship non-sage figures, including scholars, officials, monarchs, moral models, warriors, and folk gods. There are a total of 3,172 non-sage temples, covering a larger geographic area than the sage temples. By comparing the economic impacts of the number of sage temples and the number of non-sage temples, I can estimate the effect of classical worship net of the other forms of worship and the correlates of temple construction.

Modern Economic Growth

Given that China’s early economic modernization centered on industrialization, I use the number of industrial firms as the primary outcome variable. The Chinese elite began to develop new industrial firms following the Western model in the mid-nineteenth century. These new firms differed from the traditional handicraft workshops in that they relied on modern industrial technologies such as machinery, steam, and electric power, had larger numbers of employees and greater capital, and, therefore, adopted modern management systems (e.g., factory management, joint venture/stock ownership, and limited liability).

I obtained the industrial firm data from Du (Reference Du1991), who systematically recorded all domestic industrial firms established in each prefecture between 1858 and 1927. For each prefecture, I aggregate the number of firms founded during this period as the primary dependent variable. This variable gauges the cross-prefectural variation in industrial development over 70 years, covering the period of China’s early industrialization from the traditional economy.

As shown in Figure 1, the variation in the number of industrial firms across the 269 prefectures is striking. It is clear that firms preferred to locate in coastal and economically prosperous areas with good transport connections. As well as these concrete geographic conditions, firms needed to develop in regions where people were open to modernization, simply because the technology and business form of these firms were very new, Western in style, and in competition with traditional incumbents. Otherwise, in regions lacking a modern or Western outlook, firms would find it difficult to gain financing and support from the local elite (Rawski Reference Rawski1989; Du Reference Du1991)Footnote ³; they would also face a shortage of modern human capital, in particular engineers, technicians, and management professionals who play pivotal roles in operating modern industrial firms (Yuchtman Reference Yuchtman2017).Footnote ⁴

Control Variables

Basic Patterns

To examine how classical worship affected industrial development, I employ the empirical strategies specified in the following equations:

(1) $$Pre - Industria{l_i} = \alpha + \beta \times SageTemple{s_i} + {\beta _2} \times NonSageTemple{s_i} + \gamma \times {X_i} + {\varepsilon _i},$$

(2) $$Industria{l_i} = \alpha + {\beta _1} \times SageTemple{s_i} + {\beta _2} \times NonSageTemple{s_i} + \gamma \times {X_i} + {\varepsilon _i}.$$

I first examine the role of sage temples in the pre-industrial economy to check whether prefectures with stronger classical worship were already more economically backward before industrialization started (Equation (1)). I employ three measures of pre-industrial economic development. The first is the population size in 1580 and 1820, which is a conventional proxy for pre-industrial economic prosperity based on the assumption that only prosperous regions could support a large population.Footnote ⁷ The second pertains to agricultural income, proxied by the agricultural tax revenue levied in each prefecture in 1820. The third measure relates to the traditional (handicraft) industry, for which I use the distribution of silk textile centers in the Ming and Qing dynasties, since silk was a representative and leading industry in traditional China. Then, I examine the effect of sage temples on the number of industrial firms between 1858 and 1927 (Equation (2)), and compare the impact of sage temples on industrial firms to their impact on pre-industrial economic measures in Equation (1).

In the analysis noted previously, I also compare the effect of sage temples to that of non-sage temples. I include non-sage temples to help disentangle the effect of classical worship from that of other forms of worship, folk religions, or correlates of temple construction. In sum, pre-industrial prosperity and non-sage temples provide two placebos to contrast the effect of classical worship on modern growth.

In both equations, X _i includes the set of geographic controls introduced in the previous section, and ε _i is the error term. To account for spatial autocorrelation that may possibly arise from the worship culture spilling over to nearby prefectures, I report standard errors clustered over a broader geographic scale based on the method of Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019). Specifically, I cluster standard errors for a radius of 136 km from the prefectural centroid, 136 km being the average distance between contiguous prefectural centroids in the sample and thus approximately covering the neighboring prefectures.

Figure 2 reports the estimation results of the two equations. The number of sage temples is positively correlated with population size, agricultural tax revenue, and the probability of having a silk textile center in traditional China. These results indicate that classical worship was more prevalent in economically prosperous areas of pre-industrial China. This is because only developed regions could afford temple construction, classical education, and consequently, a classical intellectual culture.

Figure 2 THE ROLE OF CLASSICAL WORSHIP IN PRE-INDUSTRIAL AND INDUSTRIAL DEVELOPMENT

Notes: Population in 1580 and 1820 (log), agricultural tax in 1820 (log), and silk textile center dummy measure pre-industrial economic prosperity, and the number of industrial firms established in 1858–1927 reflects the progress of industrialization across the 269 Chinese prefectures. The number of sage temples measures the strength of classical worship. The number of non-sage temples is used as the placebo; these include all temples that worship non-sage figures (scholars, officials, monarchs, warriors, moral models, and folk gods) in each prefecture. All are OLS estimates except for the Probit estimates for the silk textile dummy. The coefficients (with 95 percent conference intervals) of the sage and non-sage temples are estimated in a single regression (Equations (1) and (2)) conditional on the log distance to the coast, log distance to the nearest navigable river, terrain ruggedness index, log land area, and agricultural suitability index (for planting wheat and rice). The actual coefficients of the sage- and non-sage temples are close to zero in the four columns for pre-industrial economic prosperity. To fully show these coefficients and confidence intervals in the figure, I multiply them by 10. Standard errors are clustered within a radius of 136 km to account for possible spatial correlation between contiguous prefectures, based on Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019) in all OLS regressions. The 136 km is the average distance between the centroids of all the contiguous prefectures in the sample. Robust standard errors are used in the Probit estimates for the silk textile dummy.

Source: See the text.

However, after industrialization began, sage temples were negatively associated with the number of industrial firms: one additional sage temple reduced the number of industrial firms by nearly four (a decrease of 35 percent evaluated by the mean number of industrial firms). This implies that classical worship may have had a hampering effect on industrial development. This effect is further highlighted by comparing it to the effect of non-sage temples, which remains positive for both pre-industrial and industrial development.

The results displayed in Figure 2 also indicate that modern industries did not necessarily emerge in traditionally prosperous regions, but in those with fewer constraints from classicism (all else equal). Next, I consider other prefectural factors that may affect both classical worship and industrial development to determine whether the negative effect of classical worship holds.

Additional Controls

To address the possible effect of pre-industrial economic conditions on industrial development, I control for each prefecture’s population in 1820. I also control for the distribution of treaty ports to gauge the impact of Western influence. To isolate the effect of state capacity, I control for the number of counties and the number of post offices in each prefecture. To disentangle classical worship from classical human capital (and its rewards from the imperial examination), I control for each prefecture’s examination quota and number of degree holders. The ordinary least squares (OLS) estimation shows that these factors do not mitigate the negative effect of sage temples on industrial development (Column (2) of Table 1). Compared to the results without these additional controls (Column (1)), the effect of sage temples becomes even greater. Since the number of industrial firms has count data and a large share of zeros (Online Appendix 2, Figure A4(b)), I also use Poisson regression and find that the effect of sage temples still holds (Column (3) of Table 1).

Table 1 CLASSICAL WORSHIP AND INDUSTRIAL DEVELOPMENT: ADDITIONAL CONTROLS

* = Significant at the 10 percent level.

** = Significant at the 5 percent level.

*** = Significant at the 1 percent level.

Notes: The table reports the cross-prefectural estimates of Equation (2) with additional controls on economic conditions, state capacity, and classical human capital. Economic conditions include pre-industrial prosperity measured by the population in 1820, and a treaty port dummy indicating whether a prefecture had a treaty port by 1927. State capacity is proxied by the number of counties and the number of post offices. Classical human capital is measured by the number of degree holders (jinshi) in the imperial examination born in each prefecture between 1371 and 1820 and the examination quota assigned to each prefecture by 1905. Column (4) examines the effect of classical worship after the rewards from classical education/examination were gone. Geographical controls include the log distance to the coast, log distance to the nearest navigable river, terrain ruggedness index, log land area, and agricultural suitability index for planting staple crops (wheat and rice). Standard errors in (parentheses) are clustered within a radius of 136 km to account for possible spatial correlation among neighboring prefectures based on Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019). Robust standard errors are in [brackets].

Source: See the text.

The payoffs of retaining classicism were closely associated with the classical institutions of imperial China, chiefly the education and examination system that based students’ career mobility on their efforts in sage learning. However, China began to modernize its education system in 1901 and then abolished the examinations in 1905, removing the institutional foundation of classical worship. Therefore, the rewards from studying classics sharply declined after 1905.Footnote ⁸ If sage temples still influenced the growth of industrial firms during this period, this would be more likely to reflect a cultural effect from intellectual conservatism (rather than the effect of classical education institutions). When I restrict the sample to the post-1905 period, sage temples still have a significant negative effect on industrial firms, suggesting that the traditional intellectual culture persisted after formal institutions that promoted this culture closed (Column (4) of Table 1).

A related concern is the trend of the classicism effect. Was the negative effect of classical worship on the number of industrial firms gradually mitigated by the sustained modernization that took place during the 70-year study period, or was it driven by variations in firm establishment in particular years? To address these questions, I examine the dynamic effect of sage temples on the number of firms over time. I aggregate the number of firms by seven decadal intervals between 1858 and 1927 for each prefecture and regress it on the interaction terms between the number of sage temples and the seven decadal dummies. Figure A5 in Online Appendix 2 displays the trend of the effect of sage temples. The number of sage temples has a persistently negative impact on the number of industrial firms, which indicates that regions with a burden of classical history gradually fell behind the others in industrialization. The non-sage (placebo) temples have no negative impact on industrial firms during this period.

Decomposing the Worship

To further investigate the temple effect from the worship of past wisdom, I divided the temples into three groups based on when the sages enshrined in them lived: (1) ancient times (580–221 BCE), (2) early to middle imperial period (221 BCE–960), and (3) later imperial times until 1820. These three periods vary in the achievements and degree of influence of sage learning. The ancient period was when most sages lived and wrote classic works. The systematization and exegesis of the ancient classic letters occurred in the second period. In the third period, scholars continued to do the exegesis work, but there was significant development and “reformation” of the classics, notably the development of the neo-Confucian School in the Song Dynasty (960–1279) and the School of Mind in the Ming (1368–1643), among others (Bol Reference Bol2008). Therefore, if classical worship did matter, worshipping the more ancient (and perhaps more sacred) wisdom should have had a greater (negative) impact on modern growth.

I run a horse race between the three groups of sage temples, controlling for the same set of covariates as in Column (2) of Table 1, and plot the results in Figure A6(a) of Online Appendix 2. The results are consistent with the hypothesis noted previously. Temples built to worship ancient sages have the greatest negative relationship with the number of industrial firms, whereas those that honor more recent masters have a smaller effect, in both magnitude and significance.

Next, I decompose the non-sage temples into six types based on the figures they enshrined (scholars (but not sages), officials, monarchs, warriors, moral exemplars, and folk gods)Footnote ⁹ to examine whether some specific types of worship—such as worship of learning (scholars) or power (officials and monarchs)—may confound the effect of classical worship on industrial development. I horse-race the sage temples with different types of non-sage temples in the same regression, and find that the negative effect of sage temples remains robust (Online Appendix Figure A6(b)).

Alternative Measures of Modern Growth

For robustness, I examine how sage temples affect alternative measures of industrial development (results summarized in Online Appendix 1, Tables A2 and A3). First, the number of firms has extreme values. While most prefectures have less than 10 firms, some have more than 100; Songjiang Prefecture or Shanghai has the most, at 735. To mitigate the impact of extreme values without trimming the sample, I winsorized the top 1 percent in the distribution of the number of firms. This does not affect the significance of the sage temple effect.

Second, I exclude prefectures that had no industrial firms during the sample period to restrict the sample to more homogenous (industrializing) regions. The number of sage temples still has a significant negative impact on the number of industrial firms.

Third, I use the per capita measure of industrial firms, normalizing the number of firms by the average population in 1910 and 1920 (per million) in each prefecture. This can mitigate the possible effect of population growth on industrial development. To mitigate the impact of extreme values, I winsorized the top 1 percent in the distribution of the per capita number of firms. The result demonstrates that sage temples still negatively affect per capita firms. The result holds when I restrict the sample to the post-1905 period (Column (4)).Footnote ¹⁰

Fourth, the number of sage temples also reduced the number of modern banks—another pillar of economic modernization that is closely associated with industrialization in China. I use the number of modern Chinese banks in 1927 (Lin et al. Reference Lin, Ma, Sun and Xu2021). One additional sage temple is associated with a 0.855 reduction in the number of banks, which indicates that the effect of classical worship spills over to the financial market.

Finally, the number of firms may not accurately measure industrial development across prefectures. Modern capital-intensive industrial firms or firms in heavy industries, for example, tend to be large to achieve economies of scale, but their number may be small. In the absence of capital and employee data, I divide firms by industrial sector to approximately mitigate the effect of industrial type on the number of firms. Given the small number of industrial firms in each sector, I aggregated some sectors that are similar in scale and capital intensity, and classified the firms into five sectors based on Du (Reference Du1991): (1) textiles, (2) food processing, (3) chemical industry, (4) machinery, transport, and mining, and (5) construction, water supply, and electricity. The results show that the negative effect of sage temples remains consistent across all industrial sectors. The textile industry, the leading manufacturing industry in China at the time (Liu Reference Liu2020), received the greatest influence from classical worship (Online Appendix 1, Table A3).

Sages and Classical Worship: Instrumented Evidence

The control variables in the analyses noted earlier may not rule out all possible factors that may confound the effect of classical worship. To address this concern, I employ an instrumental variable that predicts the strength of classical worship but has little to do with industrialization after the mid-nineteenth century—the number of sages born in each prefecture.

The positive relationship between the number of sages and the strength of classical worship is premised on the glory of the classical achievements the sages brought to their native places. In sages’ home prefectures, scholars had a greater incentive to build temples to honor the sages than in places without such historical giants. By building the sage temples, scholars also demonstrated and consolidated their localities’ orthodox status on the empire’s intellectual map (Yang Reference Yang1997).Footnote ¹¹ For example, Yanzhou Prefecture (home of Confucius and his disciples) and Jianning (home of Zhu Xi, the master of the Neo-Confucian school) maintained seven sage temples each by 1820, much more than prefectures without such sage history (on average, 0.88 temples per prefecture). Figure 3 shows the 48 home prefectures of 140 sages; it indicates that more temples were built in these holy lands.

Figure 3 DISTRIBUTION OF THE SAGES AND THEIR TEMPLES

Notes: Sages refer to the acknowledged masters of classical learning who lived between 580 BCE and 1630; before 1820, the imperial authorities formally conferred their sanctity. Sage temples are the same as those depicted in Figure 1. The map includes 269 sample prefectures of China proper based on the Qing Dynasty administration in 1820 (CHGIS, version 6).

Source: See the text.

I test the first-stage relationship in Columns (1) and (2) of Table A4 in Online Appendix 1 (see also Figure A7 in Online Appendix 2 for the visualized relationship). Regressing the number of sage temples by the number of sages yields a significant and positive link. On average, one additional sage would bring 0.147 (or 12 percent) more sage temples after controlling for population, degree holders, and geographic factors. The results also indicate that sage temples are more likely to be built in (traditionally) prosperous regions, in other words, those with a denser population and more scholars.

The placebo test on non-sage temples reaffirms the importance of the sages in shaping the geography of classicism (Columns (3) and (4) of Table A4). The number of non-sage temples is also positively correlated with economic and educational prosperity, but does not depend on the number of sages. This indicates that sages’ influence on classical worship works through the persistence of the glory of classical achievements rather than through other popular cultures and local correlates of temple building.

The sage distribution is arguably orthogonal to the industrial geography from 1858 to 1927, other than through the persistence of classical worship. This is mainly because the sages lived in the distant past, when the socioeconomic conditions of intellectual success largely differed from those during China’s modernization period. The vast majority (92 percent) of the sages were born before the thirteenth century; 73 percent of these were born in ancient times (before 256 BCE).Footnote ¹²

However, if the sages’ places of origin were economically or geographically backward, these disadvantages may persistently inhibit economic development. This is unlikely to be the case, since scholarship hardly prospered in remote or backward areas, especially in pre-industrial times. Most ancient sages were produced in the fountainhead of Chinese civilization―the North China Plain, where the earliest city-states were found. I regress a set of geographical factors on the number of sages and find that the sages were not concentrated in prefectures distant from the coast and navigable rivers or in rugged or barren lands. They tended to be born in prosperous (populous) areas (see Columns (1)–(5) of Table A5, Online Appendix 1).

Moreover, the intellectual geography changed substantially from the sage era to the nineteenth century, which is illustrated by the lack of correlation between the distribution of sages and the distribution of jinshi degree holders in the fourteenth to nineteenth centuries (Column (6) of Table A5). Therefore, the sages may not affect industrialization by shaping the distribution of classical human capital and the elite. The sages also had little impact on the popularity of clans—the lineage organizations affected by Confucianism that competed with modern economic institutions (Chen, Ma, and Sinclair Reference Chen, Ma and Sinclair2022)—by the eve of industrialization (Column (7)).

Finally, the placebo tests in Columns (5)–(7) of Table A4 provide further (suggestive) evidence of the validity of the instrumental variable. The reduced-form regression shows that the number of sages has a significant and negative effect on the number of industrial firms. This effect remains after controlling for the number of non-sage temples, but disappears after controlling for the number of sage temples in the regression. The effect of sage temples on industrial firms, though endogenous, remains significantly negative. The comparison of Columns (5) to (7) implies that sages’ long-term influence on industrialization mainly works through classical worship. Since these tests may not exclude every factor that may channel sages’ effect on industrialization beyond classical worship, we should interpret the instrumented results as suggestive rather than conclusive.

Table 2 reports the two-stage least squares (2SLS) estimates of the effect of classical worship on industrial firms. Instrumented by the number of sages, the number of sage temples still negatively impacts the number of industrial firms. The effect remains stable when controlling for geographic factors alone (Column (1)), when further controlling for economic conditions, state capacity, and classical education (Column (2)), and when the sample is restricted to the post-1905 period of educational modernization (Column (3)). Now, an additional sage temple decreases the number of industrial firms by 3.1 in the full sample and 2.9 in the post-1905 sample. The results remain robust to using the winsorized number of firms and sages to attenuate extreme values, excluding prefectures without firms, normalizing firms by population, and using modern banks as an alternative dependent variable (see Panel B of Table A2, Online Appendix 1).

Table 2 USING THE NUMBER OF SAGES TO INSTRUMENT THE EFFECT OF CLASSICAL WORSHIP ON INDUSTRIAL DEVELOPMENT

* = Significant at the 10 percent level.

** = Significant at the 5 percent level.

*** = Significant at the 1 percent level.

Notes: The table reports the two-stage least squares (2SLS) estimates at the prefectural level. Sage temples are instrumented by the number of sages. Controls are the same as those in Table 1. Standard errors are clustered within a radius of 136 km to account for possible spatial correlation among neighboring prefectures, based on Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019).

Source: See the text.

There are two remaining concerns. First, people in the holy lands of Chinese classics may have a distinctly stronger cultural triumphalism and, consequently, ethos of anti-Westernization. It might be the anti-Westernization ethos rather than the worship of ancient wisdom that inhibited the adoption of Western industrial technology. This might not be a serious concern, in view of the fact that the sage homelands did not hamper the diffusion of Christianity―the primary Western culture diffused in China at the time (Column (8) of Table A5, Online Appendix 1). Specifically, I use the number of Chinese Protestant communicants (in log) to be the proxy for Christian penetration in China, reasoning that the number of Chinese Christians would be smaller in prefectures where there was a stronger ethos against Western culture.Footnote ¹³ Anyway, given the positive impact of Christianity on China’s industrial development in the early twentieth century (Bai and Kung 2015), I control for the number of Chinese Christian communicants. The results are reported in Columns (1) and (2) of Table 3. Anti-Westernization does not affect the impact of sage temples on industrialization.

Table 3 USING THE NUMBER OF SAGES TO INSTRUMENT THE EFFECT OF CLASSICAL WORSHIP ON INDUSTRIAL DEVELOPMENT: ADDITIONAL CONTROLS AND SUBSAMPLE

* = Significant at the 10 percent level.

** = Significant at the 5 percent level.

*** = Significant at the 1 percent level.

Notes: The table replicates Table 2, but with additional controls and using a restricted sample. Columns (1) and (2) control for the number of Chinese Christians as a (inverse) proxy for anti-Westernization. Columns (3)–(6) control for the coordinates of the prefectural centroid and Reference Skinner, Henderson and YueSkinner, Henderson, and Yue’s (2013) macroregion fixed-effects. Columns (5) and (6) use the first principal component of geographic controls and that of the controls on economic conditions, state capacity, and classical education to reduce multicollinearity. Columns (7) and (8) restrict the sample to the main sage clustering region, the territory of ancient states (circa 350 BCE), where most sages were born. Standard errors in (parentheses) are clustered within a radius of 136 km to account for possible spatial correlation among neighboring prefectures, based on Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019). Standard errors in [bracket] are clustered within macroregions.

Source: See the text.

The second relates to the effect of regional clusters of sages. Many sages were born in clusters of contiguous prefectures. The most notable cluster is in the North China Plain―the cradle of Chinese civilization, where the earliest states were founded (see Figure A9 in Online Appendix 2 for the locations of ancient states around 350 BCE). This raises a concern that the effect of the instrumental variable is driven by an overall sage culture over a large region rather than by the sages of the local prefecture. This is unlikely to be the case, however, in the light of the evidence that the number of industrial firms in a prefecture is not affected by the number of sages born within 400 kilometers of the prefectural centroid, a radius that includes a prefecture’s neighbors’ neighbors (Online Appendix 1, Table A6). A possible reason is that the size of a Chinese prefecture is large enough (16,149 km² on average, six times a U.S. county of 2,465 km² on average) to dilute the impact of the neighboring prefectures.

Nonetheless, I employed two strategies to address the previously noted concern further. First, I control for each prefecture’s coordinates (at the centroid) to “fix” the spatial location effect across prefectures, and further control for macroregion dummies to fix the sage cluster effect. Skinner, Henderson, and Yue (Reference Skinner, Henderson and Yue2013) divide China into 18 macroregions based on their physiographic differences in the nineteenth century (Online Appendix 2, Figure A10). Physiographic boundaries also divide the regional cultures and socioeconomic characteristics of traditional China. This allows us to approximate the regional clusters of sage culture. However, given that both the prefectural coordinates and the physiographic regions are highly correlated with most of the other control variables, the estimate of the sage effect is attenuated by multicollinearity.Footnote ¹⁴ To address this concern, I remove all the other controls when examining the impacts of the coordinates and macroregions (Columns (3) and (4) of Table 3). Then I control for the first principal component of the geographic controls and that of the variables for economic conditions, state capacity, and classical education to reduce multicollinearity (Columns (5) and (6)).Footnote ¹⁵ In both cases, I cluster standard errors at the macroregion level to address the possible autocorrelation within the regions. The results show that, either with the other controls or without, the coordinates and the macroregions do not mitigate the effect of classical worship.

Second, I restrict the sample to the 132 prefectures that fall within the territory of the ancient states―the major cluster of sages in China (Figure A9). I examine the cross-prefectural sage effect within this territory. If it was the influence of the sage cluster in this region that drove the sage effect in the whole sample of China proper, we should not observe a significant sage effect at the prefectural level within this region. However, even in this small sample, the number of sages still significantly predicts the effect of classical worship on industrial firms, suggesting that the sage effect mainly works at the prefectural level (Columns (7) and (8) of Table 3).

Modern Journals

Modern journals were introduced to China in the late nineteenth century and became a primary medium for introducing modern sciences, technologies, and other new subjects. The first modern journal, Peking Magazine, was founded by the missionaries William Martin and Joseph Edkins in Peking in 1872. It inspired Chinese elites to develop journals to pursue new knowledge and modernization. From 1872 to 1927, they founded 424 journals that published essays on science and technology in 38 cities, according to the statistics in Yao, Wang, and Yao (Reference Yao, Wang and Yao2008).Footnote ¹⁶ I counted the number of journal editorial offices located in each prefecture. If a prefecture’s intellectuals retained stronger classical beliefs, they should have been less likely to sponsor modern journals.

To test this assumption, I regress the number of modern journals on the number of sage temples at the prefectural level, controlling for the same set of prefectural covariates as in Table 1. The result shows that there are significantly fewer journals in prefectures with more sage temples (Table 4). This result remains robust to using the Poisson estimation to address the large share of zero observations, instrumenting for classical worship using the number of sages, or examining only the journals founded after classical education was abolished. The instrumented results indicate that one additional sage temple is associated with 1.1 fewer journals. This implies that intellectual conservatism hampered the diffusion of new knowledge. In other words, modern knowledge found a small market of ideas in regions dominated by classicism.

Table 4 CLASSICAL WORSHIP AND MODERN JOURNALS

* = Significant at the 10 percent level.

** = Significant at the 5 percent level.

*** = Significant at the 1 percent level.

Notes: Modern journals refer to the number of journals on science and technology founded in each prefecture between 1872 and 1927. Sage temples are predicted by the number of sages in Columns (3) and (4). Controls are the same as those in Table 1. Standard errors in (parentheses) are clustered within a radius of 136 km to account for possible spatial correlation among neighboring prefectures, based on Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019). Robust standard errors are in [brackets].

Source: See the text.

Modern Human Capital

Prior research has recognized the importance of engineering human capital for industrialization (Mokyr Reference Mokyr2005; Maloney and Valencia Caicedo Reference Maloney and Valencia Caicedo2022). China experienced a growth in engineering talent in the late nineteenth century. Ziyuan Weiyuan Hui’s Who’s Who Chinese Engineers (1941) contains biographical information on 10,723 engineers born between 1880 and 1920, including all recorded Chinese graduates of engineering schools and some eminent engineering practitioners. I exclude the 5,071 engineers recorded in the book that lack complete information from the sample.

The engineers’ birthplaces are distributed in 85 percent of Chinese prefectures; the numbers range from 1 to 901. By 1920, about 60 years after industrialization started in China, 39 prefectures still had not produced any engineers, and nearly half had produced fewer than 10, even though some were economically prosperous or close to the country’s industrial centers. This implies that China’s industrial development did not necessarily convince local students to study engineering or industrial technologies. In other words, some regions may have lacked sufficient “push” forces in this respect. In this subsection, I test whether this was partly due to the historical burden of classicism.

I regress the density of engineers born in each prefecture from 1880 to 1920 on the number of sage temples (Table 5, Columns (1) and (2)). The number of engineers was normalized by a million residents in 1910. To normalize the distribution, I take the natural logarithm of the density (plus 1 to keep 0 values). The OLS and instrumented estimations demonstrate that sage temples have a significant and negative impact on the density of engineers. Using the number of sages to instrument the number of sage temples, one additional temple reduces the number of engineers by about 24 percent (or six engineers per million people when evaluated by the mean). Since the engineers’ distribution is based on their home prefectures, the results imply that students from regions with a stronger classicism background were less likely to choose engineering professions.

Table 5 CLASSICAL WORSHIP AND MODERN HUMAN CAPITAL

* = Significant at the 10 percent level.

** = Significant at the 5 percent level.

*** = Significant at the 1 percent level.

Notes: Engineers refers to the number of engineering school graduates and notable engineering practitioners born in each prefecture from 1880 to 1920, normalized by million inhabitants in 1910. Modern school enrollment denotes the average number of students enrolled in modern primary and middle schools in 1907, 1909, and 1914 in each prefecture (normalized by a million inhabitants in 1910). Sage temples are predicted by the number of sages in Columns (2) and (4). Controls are the same as those in Table 1. Standard errors in (parentheses) are clustered within a radius of 136 km to account for possible spatial correlation among neighboring prefectures, based on Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019).

Source: See the text.

To assess the broader effect of classical worship on human capital modernization, I measure the latter using the number of students enrolled in modern primary and secondary schools.Footnote ¹⁷ The Qing Dynasty began modernizing the education system under the New Policies Reform in 1901. New schools were gradually established that adopted modern curricula and statutes following Western models. During the eight years of primary and five years of secondary schooling, students in these new schools were required to take STEM subjects, foreign languages, and geography in addition to traditional subjects such as literature and history. There were also elective courses on law, finance, and crafts. The subsequent Republican regime continued this educational modernization.

The national education surveys conducted by the Imperial Educational Ministry (1907, 1909) and the Ministry of Education (1914) reveal the regional disparity in the progress of educational modernization. After a decade of reform, modern school enrollment increased significantly but varied from zero to more than 40,000 per million inhabitants across prefectures. Therefore, although the state promoted the reform, the effort and pace of modernizing education varied strikingly across regions.

To test whether this uneven progress was partly driven by the geography of classicism, I regress modern school enrollment on the number of sage temples, controlling for the same set of covariates as in Column (2) of Table 1. I use the average number of modern students in the three survey years (1907, 1909, and 1914) and normalize it by per million inhabitants in 1910. To normalize the distribution, I take the natural logarithm of the number. The results in Table 5 (Columns (3) and (4)) indicate that sage temples have a significant and negative impact on enrollment in modern schools. Using the number of sages to instrument for the number of sage temples, an additional temple would reduce enrollment by about 19 percent.Footnote ¹⁸

In sum, the results reported in Table 5 indicate that classical worship may have a talent allocation effect. More students chose to take courses in the new curricula and pursue industrial occupations in regions with fewer classical constraints. This pattern coincides with the tardy adoption of technical curriculum (but the persistence of religious education) in the more Catholic regions of late nineteenth-century France (Squicciarini Reference Squicciarini2020). Likewise, theological students and occupations declined in Protestant regions after the Reformation in Europe, yet students were reluctant to pursue new or secular occupations in Catholic-dominant regions (Cantoni, Dittmar, and Yuchtman Reference Cantoni, Dittmar and Yuchtman2018).

The Correlation between Modern Human Capital and Modern Growth

Finally, I illustrate the close relationship between modern knowledge and human capital, on the one hand, and China’s industrial development, on the other. I regress the number of industrial firms on the measures of modern knowledge and human capital discussed earlier. I use the number of industrial firms established between 1906 and 1927, a period after classical education was abolished that coincides with the period of modern human capital measures.

Table 6 reports the regression results. The number of industrial firms is highly correlated with the number of modern science and technology journals (Columns (1) and (2)). One additional journal published in a prefecture is associated with an increase of about 3.5 industrial firms. This calculation only captures the correlation between the journals and industrial development in local prefectures; it does not measure the possible impact of journals spilling over to other prefectures through subscriptions.

Table 6 CORRELATION BETWEEN MODERN HUMAN CAPITAL AND INDUSTRIAL DEVELOPMENT

* = Significant at the 10 percent level.

** = Significant at the 5 percent level.

*** = Significant at the 1 percent level.

Notes: Controls are the same as those in Column (2) of Table 1. Standard errors in (parentheses) are clustered within a radius of 136 km to account for possible spatial correlation among neighboring prefectures, based on Colella et al. (Reference Colella, Lalive, Orcan Sakalli and Thoenig2019). Robust standard errors are in [brackets].

Source: See the text.

There are also significantly more industrial firms in prefectures in which more engineers were born (Columns (3) and (4) of Table 6). This is consistent with Yuchtman’s (2017) finding that graduates of science and engineering schools played a more significant role in China’s modern industry than those who received a classical education. However, while the number of students in modern (primary and secondary) schools is positively correlated with the number of industrial firms, the correlation is not statistically significant. This finding coincides with the dominant role of elite (rather than common) human capital in determining industrial growth in nineteenth-century Europe (Mokyr Reference Mokyr2005; Squicciarini and Voigtländer Reference Squicciarini and Voigtländer2015) and the Americas (Maloney and Valencia Caicedo Reference Maloney and Valencia Caicedo2022). In sum, the results in Columns (3) and (4) suggest the leading role of modern “upper-tail” human capital in China’s early industrialization.