Introduction

This is a continuation of our series devoted to Carl Crow’s Newspaper Directory of China. In this new installment, we will analyze the length, the format and the layout of periodicals. Our objective is to understand how these material features related to the language, the frequency and the place of publication. Ultimately, this trilogy will lay the foundations for a material history of the press in modern China.

For this research, we rely on the following dataset:

In the table above, each row corresponds to a unique periodical. For each periodical, we provide the following information:

Object Information Label Description Process
Volume Number of pages Page_nbr_min Minimum number of pages Given
Page_nbr_max Maximum number of pages Given
Page_nbr_variation Whether the number of pages was fixed or vary between issues Produced
Page_nbr_avrg Average number of pages Produced
Format

Size of page Page_x Page first dimension (horizontal) (inches) Given
Page_y Page second dimension (horizontal) (inches) Given
Page_ratio_x_y Ratio between page dimensions Produced
Page_surface Page surface (square inches) Produced
Layout Number of columns Col_nbr_min Minimum number of columns Given
Col_nbr_max Maximum number of columns Given
Col_nbr_variation Whether the number of columns was fixed or vary between issues Produced
Col_nbr_avrg Average number of columns Produced
Size of columns Col_x_min Minimum width of column (inches) Given
Col_x_max Minimum width of column (inches) Given
Col_x_variation Whether the width of column was fixed or vary between issues Produced
Col_x_avrg Average width of column (inches) Produced
Col_y_min Minimum height of column (inches or ems) Given
Col_y_max Maximum height of column (inches or ems) Given
Col_y_variation Whether the height of column was fixed or vary between issues Produced
Col_y_avrg Average height of column (inches or ems) Produced


While most information consists in “raw” data drawn directly from the directories, some have been produced to facilitate historical analyses. In the above table, the first column indicates the chief object of inquiry (volume, format, layout). The three next columns label and describe the related information. The last column specifies where they come from and how they were produced.

Let’s start with one of the most palpable features - the length or volume of periodicals, chiefly defined by the number of pages.

Volume

Number of pages (General)

The histograms reveal huge discrepancies between periodicals. The number of pages ranges from 2 to 1000 in 1931, and from 2 to 1200 in 1935, with a slightly decreasing average of 31 in 1931, and 27 in 1937. Information is missing for three periodicals in 1931 (two quarterlies and one weekly) and two in 1935 (two weeklies).

The interactive histograms are enriched with statistics which you can explore by zooming in and compare the figures between the two directories:

Detailed statistics in 1931:

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##       2       4       8      31      16    1000       3

Detailed statistics in 1935:

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##       2       4       8      27      14    1200       3

There was a clear, reverse correlation between the number of pages and the number of periodicals. The larger the number of pages, the fewer the periodicals. Periodicals with less than 10 pages represent a clear majority in the two directories - 60% in 1931 (213) and 67% in 1935 (467). The next group with 10 to 25 pages amounted to 19% (66) in 1931 and 14% (97) in 1935. Periodicals containing from 25 to 50 pages represented less than 10% - 8% (29) in 1931, and 9% (63) in 1935. Those that included 50 to 100 pages were even fewer and declined between the two editions, even though they increased in number - from 26 (7%) in 1931, to 34 (5%) in 1935. More voluminous publications that contained 100 to 500 pages maintained their position (5%).

The directories listed just four periodicals with more than 500 pages. In 1931, they included two annuals - one in English, the Hongkong-based Directory and Chronicle of China, Japan, Straits, Philippines, etc. (1200p), the other in Chinese, the Commercial Directory of Shanghai 上海商業名錄 (Shanghai shangye minglu) (780p) - and two Chinese-language monthly journals - Current Events 時事月報 (Shishi yuebao) and Public Education Monthly 民衆敎育月刊 (Minzhong jiaoyu yuekan) with 600 pages each. In 1935, there were just annuals and only the Commercial Directory of Shanghai remained. The three new giants were three reference directories in contemporary business circles: The Credit Men’s Business Directory of China 徵信工商行名錄 (Zhengxin gongshanghang minglu) (1200p), the Shanghai Dollar directory (900p) and the China Hong List (880p).

Most voluminous periodicals (more than 500 pages) in 1931:

Most voluminous periodicals (more than 500 pages) in 1935:

To sum up, small-volume periodicals, with the smallest number of pages, represented the majority and were rapidly growing in the early 1930s, at the expenses of intermediate volumes (especially those between 10 to 25 pages), while more voluminous publications enjoyed greater stability. We suspect the growing small-volume periodicals refer to the most frequent ones - the dailies, tabloids and weeklies we have shown the growth in a previous essay - whereas more voluminous, stable publications, correspond to periodicals with less frequent periodicity. The next section examines this hypothesis in greater depth.

Periodicity

How did the number of pages relate to frequency of publication?

In order to modelize the relation between the volume and the frequency of publication, we performed a basic linear regression analysis (lm) with the number of pages as the outcome variable (y) and the periodicity as the explanatory variable, setting daily newspapers as the reference category.

The model reveals a strong correlation in the two directories (p value <2e-16). Volume was clearly a factor of differentiation between periodicals with different periodicity. The most frequently published (dailies, tabloids, weeklies) contained fewer pages than monthlies, quarterlies and annuals. The prevalence of small-volume periodicals essentially reflects the domination and growth of the daily press in early 1930s China, as we showed in the previous essay.

Correlation between the number of pages and the frequency of publication in 1931:

## 
## Call:
## lm(formula = Page_nbr_avrg ~ Periodicity, data = crow_format31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -543.33   -4.38   -0.38    3.62  482.78 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             8.381      3.652   2.295  0.02233 *  
## PeriodicityAnnual     609.953     32.798  18.597  < 2e-16 ***
## PeriodicityBiweekly    14.119     28.462   0.496  0.62016    
## PeriodicityMonthly    108.837      9.090  11.974  < 2e-16 ***
## PeriodicityQuarterly   90.786     23.335   3.891  0.00012 ***
## PeriodicityTabloid     -5.381     40.086  -0.134  0.89330    
## PeriodicityWeekly      11.565      8.443   1.370  0.17164    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 56.45 on 348 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.5806, Adjusted R-squared:  0.5734 
## F-statistic:  80.3 on 6 and 348 DF,  p-value: < 2.2e-16

All categories, except tabloids, generally contained a larger number of pages than daily newspapers. The difference with the least frequent periodicals was particularly huge. Annuals, monthlies and quarterlies generally contained 609, 108 and 90 more pages than dailies, respectively. Biweeklies and weeklies had 14 and 11 more pages than dailies, respectively. Note: The figures for biweeklies, tabloids and weeklies, however, should be taken with caution, given that the p-value for these categories is superior to 0.05.

Correlation between the number of pages and the frequency of publication in 1935:

## 
## Call:
## lm(formula = Page_nbr_avrg ~ Periodicity, data = crow_format35)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -160.000   -3.213   -0.250    2.787  260.000 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                7.213      1.304   5.532 4.50e-08 ***
## PeriodicityAnnual        932.787     13.967  66.783  < 2e-16 ***
## PeriodicityBimonthly     102.787     27.843   3.692  0.00024 ***
## PeriodicityBiweekly        2.787     13.967   0.200  0.84191    
## PeriodicityMonthly        85.099      3.372  25.238  < 2e-16 ***
## PeriodicityQuarterly     139.787      9.919  14.092  < 2e-16 ***
## PeriodicitySemi-monthly   60.787      8.134   7.473 2.38e-13 ***
## PeriodicitySemi-weekly    -2.963      7.074  -0.419  0.67545    
## PeriodicityTabloid        -3.170      4.303  -0.737  0.46161    
## PeriodicityWeekly         16.828      3.507   4.799 1.96e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 27.81 on 689 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.8832, Adjusted R-squared:  0.8817 
## F-statistic: 578.9 on 9 and 689 DF,  p-value: < 2.2e-16

The gap had widened by 1935. Annuals now exceeded dailies by 932 pages, quarterlies by 139, bimonthlies by 102, monthlies by 85 and semi-monthlies, by 60. Weeklies had 16 more pages than dailies, but biweeklies, only 2. Tabloids and semi-weeklies were the only two categories that generally contained fewer pages than dailies - 3 and 2, respectively. Note: The figures for semi-weeklies, biweeklies and tabloids should be taken with caution, given that the p-value for these categories is superior to 0.05.

The boxplots below reveal important variations in the volume of periodicals across and within categories. Annuals contained the largest number of pages in the two directories but the range decreased between the two editions. On the opposite, tabloids contained the fewest number of pages. Dailies, weeklies and intermediate periodicals (bi/semi-weeklies) also contained few pages but there may be huge variations within each category, especially weeklies. Monthlies, semi-monthlies and quarterlies stood in between.

The interactive version is enriched with statistics for each category of periodical. You can zoom in and compare the figures across periodicals and between the two directories (1931 on the right, 1935 on the left):

Let’s have a closer look at each category.

Dailies: their volume range from 2 to 26 in 1931, and 2 to 32 in 1935, with an average of 8 in 1931, in 1935. The most voluminous in 1931 include the Hong-kong based South China Morning Post, the Russian Kung Pao in Harbin, and the Chinese Xinwenbao in Shanghai (26 pages each). In 1935, the two Shanghai giants Shenbao 申報 and Xinwenbao 新聞報 competed with 32 pages each. Next we found the German Deutsch-Chinesische Nachrichten in Tianjin and three major English-language newspapers - China Press 大陸報 (Dalubao) (29p), North-China Daily News (25p) and South China Morning Post (22p).

Small-volume dailies (less than 3 pages):

Most voluminous dailies (over 25 pages in 1931, over 30 pages in 1935):

Tabloids presented greater uniformity. Their volume ranged from 2 to 16 pages, with a median of 4. The only 16-page tabloid was the Nanjing-based Press of the People 民生報 (Mingshengbao). The three most voluminous tabloids (over 10 pages) were all based in Nanjing:

The fifteen two-page tabloids were scattered across four different cities, namely Chongqing, Shanghai, Wuxi and Zhenjiang:

Biweeklies (twice a week): their volume varied from 2 to 80 pages in 1931, 4 to 28 in 1935, with a median of 4. The Changzhou-based Steel Press 鋼報 (Tiebao) was the only two-page biweekly listed in the 1931 directory:

Although its volume had doubled by 1935, it still belonged to the least voluminous periodicals in this category, along with the two other four-page biweeklies, the Foochow People’s Press 建甌民報 (Jian’ou minbao) and Singling People’s Press 新寧民報 (Xinning minbao):

The two most voluminous biweeklies included the Liaoning Educational Journal 遼甯敎育公報 (Liaoning jiaoyu gongbao) in Shenyang (80p) and The China Truth (English) in Guangzhou (28p):

Note: the figure given for the Liaoning Educational Journal seems abnormally high for a biweekly publication. The editor may have mistaken “eighty” for “eighteen” (Crow, 1931, p.19). Unfortunately, the periodical had disappeared by 1935, so it is impossible to cross-reference the information with the later edition.

The volume of weeklies volume span from 2 to 80 pages. Moreover, the median considerably increased (from 4 to 21) between the two editions. The smallest weeklies (less than 5 pages) consisted mostly in illustrated periodicals. The Sin Wen Pao Pictorial 新聞報畫報 (Xinwenbao huabao) was the only two-page weekly listed in the 1931 directory. Small-volume weeklies (less than 5 pages) consisted mostly in illustrated periodicals:

Seven weeklies exceeded fifty pages. The three most voluminous - Mongolia and Thibetan Weekly 蒙藏週刊 (Mengzang zhoukan) in Nanjing, Bankers’ Weekly 銀行週報 (Yinhang zhoubao) in Shanghai, and National News Weekly 國聞週報 (Guowen zhoubao) in Tianjin - even gained five pages between 1931 and 1935 (rising from 75 to 80 pages):

As for monthlies, their volume ranged from 4 to 600 pages in 1931, 4 to 300 in 1935, with an median of 100 pages in 1931, 80 in 1935. Monthlies were one of the rare categories (along with quarterlies) that show a decrease in the maximum (and median) number of pages. The two Nanjing-based Public Education Monthly 民衆敎育月刊 (Minzhong jiaoyu yuekan) and Current Events 時事月報 (Shishi yuebao) were the two densest monthlies in 1931 (600 pages each):

The Chinese-language Public Welfare Monthly 公義月報 (Gongyi yuebao) published in Shanghai since 1931 was the only four-page monthly listed in the two directories.

Semi-monthlies may contain from 22 to 130 pages, with a median of 54. The slightest one was the Chinese women’s magazine Women’s Voice Semi-Monthly 女聲半月刊 (Nüsheng banyuekan) (22p) and the National-Defense Forum 國防論壇 (Guofang luntan) (24p) - both published in Shanghai since 1932 and 1934, respectively:

The most substantial monhtly was the pioneering Eastern Miscellany 東方雜誌 (Dongfang zazhi), established in 1903 (130p), followed by the Semi-Monthly Economic Journal 工商半月刊 (Gongshang banyuekan) (120p) - the Chinese-language counterpart to the Chinese Economic Journal/Bulletin - and the Children’s World 兒童世界 (Ertong shijie) magazine (110p), all based in Shanghai. Such variety reveals that the length of periodicals was not determined by their content and readership. Publications for children could be as copious as technical and intellectual ones. Conversely, women’s magazines were not shorter than patriotic publications.

The size of quarterlies varied from 75 to 170, with a median of 85 in 1931, and from 50 to 400 pages (median: 110) in 1935. The three most substantial were highly specialized periodicals published by academic institutions - the Chinese Journal Of Physiology (400p), the Quarterly Journal of Economics 經濟學季刊 (Jingjixue jikan) published by the Chinese Economic Society (210p) and Agriculture Science (170p) - published by the College of Agriculture at National Central University in Nanjing (170p). The copious size of these quarterlies was strongly related to their low frequency of publication, their readers’ high level of education, and the highly specialized nature of their content.

Hongkong Naturalist, The Caduceus and The Nursing Journal of China were the three least substantial quarterlies. Moreover, their volume decreased between 1931 and 1935. They all contained 75 pages in 1931. In 1935, Hongkong Naturalist was twenty-five pages slighter, while The Caduceus and The Nursing Journal had lost thirteen and eleven pages, respectively.

Annuals were the mot substantial of all periodicals. Their volume varied between 75 to 1000 pages in 1931, and from 780 to 1200 in 1935 (median: 780 in 1931, 890 in 1935). The two thousand-page annuals were two reference directories that catered to a large readership of English and Chinese businessmen scattered across China and the Pacific - the enduring Directory and Chronicle of China (established in 1862) (1000p) and the more recent Credit Men’s Business Directory of China 徵信工商行名錄 (Zhengxin gongshanghang minglu) first published in 1933 (1200p).

The Hongkong University Union Magazine was the only 75-page annual in 1931, weighting less than ten percent of the Commercial Directory of Shanghai 上海商業名錄 (780p) that came immediately after:

The huge gap between the two annuals reflects the varying size of their respective readership. The former addressed a narrow readership of English-speaking students in Hongkong, whereas the latter catered to a large and growing business community in Shanghai - the booming business center of the country (Marie-Claire Bergère, L’Âge d’or de la bourgeoisie chinoise, Paris, Flammarion, 1986).

Other periodicals were not statistically important and their volume did not vary significantly.

Language

How did the language impact of the volume of periodicals? Due to their statistical importance, the distribution of Chinese-language periodicals reflects the general trend, foreign-language periodicals presented a more complex picture.

In order to modelize the relation between the volume and the language of publication, we performed a basic linear regression analysis (lm) with the number of pages as the outcome variable (y) and the language as the explanatory variable, setting “Chinese” as the reference language category.

We observe a strong correlation in 1935 (p value 5.184e-07) but not in 1931 (p-value: 0.3319), which suggests that the number of pages gradually emerged as a factor of differentiation between languages in the early 1930s.

Correlation between the number of pages and the language of periodicals in 1931:

## 
## Call:
## lm(formula = Page_nbr_avrg ~ Language, data = crow_format31)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
## -52.83 -23.08 -19.08 -13.08 941.17 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)         27.075      5.052   5.360 1.52e-07 ***
## LanguageBilingual   60.425     61.248   0.987   0.3245    
## LanguageEnglish     31.754     13.567   2.341   0.0198 *  
## LanguageFrench     -13.075     86.470  -0.151   0.8799    
## LanguageGerman      -9.075     86.470  -0.105   0.9165    
## LanguageJapanese   -14.742     35.601  -0.414   0.6791    
## LanguageRussian    -12.742     35.601  -0.358   0.7206    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 86.32 on 348 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.01947,    Adjusted R-squared:  0.002566 
## F-statistic: 1.152 on 6 and 348 DF,  p-value: 0.3319

Correlation between the number of pages and the language of publication in 1935:

## 
## Call:
## lm(formula = Page_nbr_avrg ~ Language, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
##  -76.87  -16.36  -14.36   -8.62 1116.13 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)         20.361      3.177   6.409 2.71e-10 ***
## LanguageBilingual   43.639     78.977   0.553    0.581    
## LanguageEnglish     63.508     10.017   6.340 4.15e-10 ***
## LanguageFrench      -6.361     78.977  -0.081    0.936    
## LanguageGerman      -1.361     55.891  -0.024    0.981    
## LanguageJapanese    -8.028     32.373  -0.248    0.804    
## LanguageRussian     -7.695     45.671  -0.168    0.866    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 78.91 on 692 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.05562,    Adjusted R-squared:  0.04743 
## F-statistic: 6.793 on 6 and 692 DF,  p-value: 5.184e-07

English and bilingual periodicals contained a larger number of pages than Chinese-language ones (63 and 43 more on average, respectively). On the opposite, Japanese, Russian, French and German periodicals had fewer pages than Chinese ones (8, 7.6, 6.3 and 1 less, respectively). These figures should be taken with caution, however, given the high p-value associated with individual languages, except English.

The boxplots show important variations in the number of pages across and within languages. English periodicals reached the maximum number of pages, but the Chinese press offered the widest range. Other languages presented little variation and were not statistically significant.

The interactive version below is enriched with statistics for each language You can zoom in and compare the figures across periodicals and between the two directories (1931 on the right, 1935 on the left):

Let’s focus on the two dominant languages, Chinese and English.

The volume of Chinese periodicals ranged from 2 to 780, with a median of 8 in 1931, 7 in 1935. The most voluminous (over 500 pages) were the annual and monthlies we presented earlier:

The least voluminous Chinese periodicals consisted mostly in daily newspapers and tabloids and a handful of weeklies, bi- and semi-weeklies, i.e. periodicals with high frequency of publication:

English periodicals ranged from 6 to 1000 in 1931, 7 to 1200 in 1935 with a median of 30 in 1931, 34 in 1935. The most voluminous was the Directory and Chronicle of China, Japan, Straits, Philippines, etc. in 1931 and the Credit Men’s Business Directory of China (Including Hongkong) in 1935.

The most substantial English periodicals were the four annuals that we presented earlier:

The least substantial (less than 10 pages) were daily newspapers that were not based in Shanghai and Hongkong:

Minor languages presented greater uniformity. Japanese periodicals contained from 10 to 14 pages, German ones may included from 8 to 30 pages in 1935, Russian from 12 to 26.

The two most substantial Japanese periodicals were the Shanghai Mainichi Shimbun and Shanghai Nichi Nichi Shimbun (14 pages each). Those published in Northern China contained fewer pages. The Seito Shimpo in Qingdao was the least substantial. The Dairen Shimbun in Shenyang (Manchuria) stood in between. The volume of Japanese periodicals did not change between the two editions:

Among Russian dailies, the most copious was the Russian Kung Pao based in Harbin (26 pages). The majority was stable between the two editions, except for the Shanghai daily The Slovo that gained two pages:

While the French daily Le Journal De Shanghai was stable (14 pages), the German Deutsch-Chinesische Nachvichten almost doubled its volume (from 18 to 30) between 1931 and 1935:

To conclude this section, the differences between languages largely reflected more structural differences based on periodicity. As we observed in the previous essay, non-daily - i.e. more copious - periodicals prevailed in the English-language press, whereas daily newspapers with fewer pages dominated in the Chinese and other foreign-language presses.

Variability

For some periodicals, the number of pages may vary from one issue to the other. Such variability reflected the editors’ propensity to publish occasional supplements, special pages and editions. It also highlighted the growing importance of advertising contents. Some periodicals devoted more than half of their volume to advertising. The China Medical Journal, for instance, contained 100 pages of reading matter against 60 pages of advertisements in 1935 (Crow, 1935, p.144). Variation in volume may also reflect editorial changes and variability in the quantity and quality of information available in each edition.

Such variability, however, tended to decrease over time, reinforcing the standardization process we observed earlier:

In the 1931 edition of the directory, we found 101 periodicals (28%) with varying volume, whereas in 1935, there represented less than 10% (63, 9%), which reveals that Chinese periodicals in the erarly 1930s became increasingly standardized and enjoyed greater stability in their format. This largely reflects the growth of the daily press that had to be printed in emergency, leaving little room for change from one day to the next.

The plots below visualize in more detail the relation between the minimum (x axis) and the maximum (y axis) number of pages. Dots that stick to the line present small variation, whereas those that diverged from the line show the greatest variability:

The interactive scatterplot below is enriched with statistics for each point (periodical):

Moreover, periodicity was a strong determining factor in volume variability:

The greatest variability was to be found among monthly journals. The number of pages in Public Education Monthly and Current Events was multiplied by a factor of five (from 200 to 1000 pages). It doubled for several other monthlies, such as Child Education 兒童教育 (100 to 200), St. John’s Review (50 to 100) or Chop Sticks (20 to 40).

Weeklies also displayed significant variation. Some of them may double their content, such as the Mongolia And Thibetan Weekly (50 to 100p), Walla Walla (32 to 76), the Hongkong Weekly Press & China Overland Trade Review (20 to 40) or the Hongkong Sunday Herald (12 to 14).

Volume and surface

How did the number of pages relate to their size?

The scatterplots below visualize the correlation between the number of pages (y) and their average surface (x) measured in square inches:

The interactive scatterplot is enriched with statistics for each point (periodical):

The plots suggest some relation between the number of pages and their size. Small-size publications (dots above the line) tended to contain a larger number of pages than large-size periodicals (dots below the line, especially those in the right-hand corner). The former group included mostly annuals and quarterlies, whereas the latter consisted mostly in daily and weekly periodicals. Monthly journals and intermediate periodicals lied in-between:

In order to modelize the relation, we run a simple linear regression analysis (lm), setting the surface of the page as the outcome variable (y) and the number of pages as the explanatory variable (x).

The relation proved statistically significant and negative in the two directories (p value: 2.429e-14 in 1931, < 2.2e-16 in 1935), which means that periodicals became less substantial as their page expanded, or put the other way, their volume augmented as their pages shrank. More precisely, they gained one page when their surface decreased by 0.4 square inches in 1931 and 0.6 in 1935.

Number of pages in relation with their size in 1931:

## 
## Call:
## lm(formula = Page_surface ~ Page_nbr_avrg, data = crow_format31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -168.77  -94.02   47.98   71.41  294.79 
## 
## Coefficients:
##                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   215.57386    5.36566  40.177  < 2e-16 ***
## Page_nbr_avrg  -0.46536    0.05849  -7.957 2.43e-14 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 95.11 on 353 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.1521, Adjusted R-squared:  0.1497 
## F-statistic: 63.31 on 1 and 353 DF,  p-value: 2.429e-14

Number of pages in relation with their size in 1935:

## 
## Call:
## lm(formula = Page_surface ~ Page_nbr_avrg, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -175.32  -84.62   36.91   76.84  420.71 
## 
## Coefficients:
##                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   207.41893    3.74538   55.38   <2e-16 ***
## Page_nbr_avrg  -0.63792    0.05865  -10.88   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 91.84 on 695 degrees of freedom
##   (5 observations deleted due to missingness)
## Multiple R-squared:  0.1455, Adjusted R-squared:  0.1442 
## F-statistic: 118.3 on 1 and 695 DF,  p-value: < 2.2e-16

The next section will explore in greater depth these issues of format and page dimensions.

Format

In the “advertising manual” that followed the preface in the last edition of the directory (1937), Crow emphasized that the size of newspapers was not uniform in early 1930s China. It varied from place to place depending of the type of press used in each city. Moreover, different presses could be used in the same city, especially in large urban centers such as Shanghai. As Crow explained:

The size of a Chinese newspapers is not uniform, as in every city there is a difference due to the type of presses in use. Big cities like Shanghai, Tientsin and Canton, where papers are printed on high speed rotary presses, the paper size also does not give the same measurement. For example, the Shun Pao [Shenbao] is 15 1/4 x 20 for a full page or 305 square inches, while the Sin Wan Pao [Xinwenbao] measures 16 X 20 or 320 square inches. Province and local papers which use common platen presses have page sizes ranging from 12 x 19 inches to 15 x 20 inches.

(Carl Crow, Inc. “Newspaper Mechanical Requirements.” In Newspaper Directory of China (Including Hongkong) & Advertising Manual, 1937: An Annual Data Book and Technical Record for All Newspapers and Periodicals in China, Shanghai: Carl Crow, Inc., 1937, p.1.)

How large were the pages of periodicals in early 1930s China? How did their format evolve over time? Do we observe a reduction in the size of pages, for practical (readers’ mobility) or economical (cost of ink and paper) reasons? Did the directories confirm the rise of compact periodicals at the expense of larger broadsheets - as demonstrated in previous scholarship (Chin, 2014)? How did the size of pages relate to the language, the frequency and the place of publication?

Surface

Let’s start with a broad overview of the distribution of periodicals based on the average surface of the page:

The histograms reveal huge discrepancies between periodicals.

Detailed statistics:

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##      24     117     258     201     283     462       3
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##      19     113     240     192     280     388       3

The average surface may vary from 24 to 462 square inches in 1931, from 19 to 388 in 1935. While the largest periodicals in 1931 had not increased (and even decreased) by 1935, the smallest, one the opposite, became smaller. The mean had fallen from 201 to 192, and the median from 258 to 240, which points to a general process of reduction in the early 1930s. Information is missing for six periodicals - three in 1931 (one Chinese weekly and two English quarterlies) and three in 1935 (one Chinese weekly and two English annuals):

The interactive histograms below are enriched with statistics which you can explore by zooming in and compare the figures between the two directories (1931 on the left, 1935 on the right:

The histograms delineate three major groups. The largest periodicals with pages measuring around 300 sq.inch constituted the largest group. Small-size periodicals with pages measuring less than 100 sq.inch. represented a minority. Medium-size periodicals (100-150 sq. inches) stood in between. As we will see later, these groups largely reflect the frequency of publication, especially the prevalence of large-size daily newspapers (cf. our previous essay).

Categorization

Based on the page surface distribution, we defined three major categories to facilitate further analyses:

Format Surface (sq inches)
Octavo 0 to 100
Compact 100 to 200
Broadsheet 200 to 400

Our classification borrows from international standards, while adapting them to the reality of the press industry in Republican China. For instance, the largest Chinese newspapers in the 1930s were smaller than British broadsheets at the same period. The largest broadsheets listed in the directories measured 17 X 22 inches (388 square inches), whereas British broadsheets typically measured 29.5 by 23.5 in (749 by 597 mm). Moreover, the vertical dimension of Chinese newspapers was usually larger than their width, whereas British broadsheets often adopted a horizontal format. Chinese broadsheets were more akin to their counterparts in the United States, which usually measured 15 inches wide (381 mm) by 22.57 inches long (578 mm). We will examine in greater detail the dimensions of pages in the next sections.

The histograms confirm the domination of large-size broadsheets and the rise of small-size periodicals, especially “compact” periodicals:

Detailed statistics:

Broadsheets increased in number but decreased in proportion (57% to 53%), even though they remained the majority. Compact periodicals (100-200 sq.inch) experienced the most remarkable growth, rising from 72 (20%) to 164 (23%). Although they stagnated around 22-23%, book-modeled (octavo) formats (under 100 sq.inch) more than doubled in number (79 to 163). To sum up, small-size periodicals clearly progressed between the two editions.

Periodicity

The barplots below provide a quick visual confirmation that the smallest periodicals were also the least frequently published, whereas the largest corresponded to the most frequent ones, especially daily newspapers.


This preliminary observation, however, deserves closer analysis.

In order to modelize the relation between the size of periodicals and the frequency of publication, we run a simple linear regression analysis (lm), with the average surface of the page as the outcome variable (y) and the periodicity as the explanatory variable, setting “daily” as the reference category.

The model reveals a strong correlation in the two directories (p value <2e-16). The size was clearly a factor of differentiation between periodicals with different periodicity. Daily newspapers contained larger pages than less frequent periodicals. Moreover, the less frequent the periodicity, the largest the difference with dailies.

Size of page and frequency of publication in 1931:

## 
## Call:
## lm(formula = Page_surface ~ Periodicity, data = crow_format31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -207.14  -14.75    9.50   30.31  341.00 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           256.014      4.105  62.373  < 2e-16 ***
## PeriodicityAnnual    -218.014     36.865  -5.914 7.99e-09 ***
## PeriodicityBiweekly  -111.717     31.992  -3.492 0.000541 ***
## PeriodicityMonthly   -203.768     10.217 -19.945  < 2e-16 ***
## PeriodicityQuarterly -222.472     26.228  -8.482 6.42e-16 ***
## PeriodicityTabloid   -135.889     45.057  -3.016 0.002750 ** 
## PeriodicityWeekly    -134.765      9.490 -14.201  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 63.45 on 348 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.6279, Adjusted R-squared:  0.6215 
## F-statistic: 97.89 on 6 and 348 DF,  p-value: < 2.2e-16

The pages of annuals, for instance, were smaller by 218 square inches in 1931, 197 in 1935. For quarterlies, the difference was also significant (222 in 1931, 195 in 1935). Biweeklies, weeklies and tabloids, which were closer in frequency, had their pages smaller by just 111, 134 and 135 square inches, respectively. In between, the pages of monthlies differed from dailies by 203 square inches.

Size of page and frequency of publication in 1935:

## 
## Call:
## lm(formula = Page_surface ~ Periodicity, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -167.93  -32.20   13.82   37.08  242.31 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)              242.925      3.101  78.342  < 2e-16 ***
## PeriodicityAnnual       -197.175     46.873  -4.207 2.94e-05 ***
## PeriodicityBimonthly    -194.925     66.216  -2.944  0.00335 ** 
## PeriodicityBiweekly     -102.519     33.217  -3.086  0.00211 ** 
## PeriodicityMonthly      -185.727      8.019 -23.161  < 2e-16 ***
## PeriodicityQuarterly    -195.519     23.590  -8.288 6.01e-16 ***
## PeriodicitySemi-monthly -182.154     19.344  -9.417  < 2e-16 ***
## PeriodicitySemi-weekly  -109.972     16.824  -6.537 1.23e-10 ***
## PeriodicityTabloid       -66.941     10.134  -6.606 7.92e-11 ***
## PeriodicityWeekly       -149.240      8.291 -18.001  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 66.14 on 689 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.5618, Adjusted R-squared:  0.5561 
## F-statistic: 98.16 on 9 and 689 DF,  p-value: < 2.2e-16

The relation did not change in 1935, except for the fact that the difference between with tabloids had decreased (their pages were just 66 square inches smaller than those of dailies). Intermediate periodicity complicated the picture, but their pages were also smaller than those of dailies, and the difference was inversely proportionate to their frequency.

Moreover, the boxplots reveal important variations in the size of periodicals across and within categories. Dailies, weeklies, biweeklies and tabloids presented the widest scope of format. Variability, however, tended to decrease between the two directories (except for tabloids that were not statistically significant in the first edition).

The interactive version of the graph is enriched with statistics for each category of periodical. You can zoom in and compare the figures across periodicals and between the two directories:

Let’s examine each category in greater detail.

Weeklies presented the widest spectrum in terms of format. Their size may vary from 24 to 362 in 1931, 19 to 336 in 1935, with a decreasing median from 117 and 70. The smallest weeklies included youth publications such as Children’s World 兒童世界 (Ertong shijie) and New Schoolmate 新同學 (Xintongxue), as if the size of periodicals should be proportionate to the body of their expected readers:

On the opposite, the largest weeklies were mostly English and Hongkong-based well established periodicals devoted to general news, such as South China Sunday News 南華星期報 (Nanhua xingqibao), Hongkong Sunday Herald and Shanghai Sunday Times:

Dailies also displayed great variety in their format. Their size ranged from 48 to 388 in 1931, 75 to 388 sq. inches in 1935. The smallest ones (under 100 sq. inches) were all Chinese newspapers published outside of Shanghai:

Among the largest dailies (over 300 sq. inches), English newspapers surpassed all other languages. The pioneer North-China Daily News 字林報 (Zilinbao) held the record (388 sq. inches). The largest Chinese newspapers were published outside of Shanghai, especially in Guangzhou, Hongkong, Shantou and Yangzhou. The list also includes the French Journal de Shanghai (361 sq. inches.):

As they multiplied between 1931 and 1935, tabloids also diversified their format. Their size may vary from 48 to 333 sq. inches in 1935 (median: 137). Clearly, we shall beware of automatically associating tabloids with “compact” formats, as previous scholarship had been inclined to (Chin, 2014).

The smallest tabloid (48 sq. inches) was the Strong Group Press 羣强報 (Qunqiangbao) published in Beijing (classified as a “daily” in the 1931 edition). The list also includes several Shanghai-based movie magazines, such as Movie Life 電影生活 (Dianying shenguo), Camera News 開麥拉 (Kaimaila) and Radio Film Daily News 電影日報 (Dianying ribao). Portability seemed to be a specific feature of leisure periodicals.

Among the largest tabloids (over 300 sq. inches), we found three dailies devoted to general or local news, such as Eastern Daily News 東方日報 (Dongfang ribao) and New Shanghai 新上海 (Xin Shanghai):

The pages of biweeklies measured between 37 and 265 sq. inch. in 1931, between 76 and 265 in 1935 (with a median decreasing from 137 to 110).

The pages of monthlies ranged from 24 and 126 sq. inch. in 1931, between 24 and 153 in 1935 (with a median decreasing from 43 to 40). Monthlies were among the rare periodicals whose maximum size increased between the two directories.

The size of semi-monthlies may vary from 24 to 93 sq. inches in 1935 (median: 58). As for weeklies, the smallest include publications for children but also more “serious” periodicals, such as the Semi-Monthly Economic Journal

Quarterlies measured from 28 to 78 (median: 38). The smallest was the bilingual medical journal Nursing Journal Of China published in Nanjing, whereas the largest ones consisted in more popular and patriotic publications, such as the Young Companion 良友 (Liangyou) (93.5 sq.inches) and the National-Defence Forum 國防論壇 (Guofang luntan) (82.5).

Other categories were not statistically significant. As suggested earlier, annuals were among the smallest of all periodicals. This reflects their publishers’ origin and their proximity with the book industry, which we will explore in a future essay.

Language

While the distribution of Chinese-language periodicals based on their format reflects the general distribution, foreign-language periodicals presented a more complex picture.

English-language periodicals were represented in all categories, but were generally of smaller size than other foreign languages. This reflects the greater variety of English periodicals (in terms of frequency), compared to other foreign presses, which consisted mostly in daily newspapers.

In order to analyze more systematically the relation between the size and the language of periodicals, we run a simple linear regression analysis (lm), with the average page surface as the outcome variable (y) and the language as the explanatory variable (x), setting “Chinese” as the reference category.

The difference was significant in the two directories (p value: 2.931e-05 in 1931, 1.224e-07 in 1935). The size was clearly a factor of differentiation between periodicals published in different languages.

Size and language of periodicals in 1931:

## 
## Call:
## lm(formula = Page_surface ~ Language, data = crow_format31)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
## -181.5  -84.0   44.5   74.5  256.8 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        205.500      5.827  35.267  < 2e-16 ***
## LanguageBilingual -173.500     70.648  -2.456 0.014543 *  
## LanguageEnglish    -53.029     15.649  -3.389 0.000783 ***
## LanguageFrench     155.750     99.741   1.562 0.119304    
## LanguageGerman      -5.500     99.741  -0.055 0.956055    
## LanguageJapanese    95.791     41.065   2.333 0.020235 *  
## LanguageRussian     92.312     41.065   2.248 0.025205 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 99.57 on 348 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.0839, Adjusted R-squared:  0.06811 
## F-statistic: 5.312 on 6 and 348 DF,  p-value: 2.931e-05

Chinese periodicals were generally larger than German, English and bilingual ones, by 5, 53 and 153 square inches, respectively. The gradation in difference mostly reflects the decreasing periodicity associated with each of these languages. Russian, Japanese and French periodicals were larger by 92, 95 and 155 square inches, respectively. These figures, however, should be taken with caution, since the p-value was superior to 0.05.

Size of page and language in 1935:

## 
## Call:
## lm(formula = Page_surface ~ Language, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -177.81  -76.86   36.98   82.97  255.69 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        197.025      3.889  50.667  < 2e-16 ***
## LanguageBilingual -169.025     96.749  -1.747  0.08107 .  
## LanguageEnglish    -64.587     12.351  -5.229 2.26e-07 ***
## LanguageFrench     164.225     96.749   1.697  0.09006 .  
## LanguageGerman      11.975     68.467   0.175  0.86121    
## LanguageJapanese   103.412     39.657   2.608  0.00931 ** 
## LanguageRussian     92.975     55.948   1.662  0.09701 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 96.67 on 692 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.05994,    Adjusted R-squared:  0.05179 
## F-statistic: 7.354 on 6 and 692 DF,  p-value: 1.224e-07

Except for German periodicals, which now surpassed Chinese ones by 11 square inches, the relation remained the same in 1935. The gap with English, Japanese and French periodicals had widened (from 53 to 64, 95 to 103, from 155 to 164, respectively). The difference with Russian newspapers remained stable. The figures for bilingual, French and Russian periodicals, however, are not statistically reliable.

Moreover, the boxplots reveal important variations in the size of periodicals across and within languages.

The interactive version of the graph is enriched with statistics for each language category. You can zoom in and compare the figures across languages and between the two directories:

The Chinese and English-language press offered the greatest variety. The size of Chinese periodicals varied between 24 to 362 sq. inches in 1931, 19 to 377 in 1935 (median: 260 in 1931, 243 in 1935). English periodicals measured from 31 to 388 sq. inches in 1931, 25 to 388 in 1935 (median: 76 in 1931, 74 in 1935). Japanese, Russian and German newspapers presented greater uniformity, with pages ranging 295 to 312-317 (median: 300), from 275 to 344 in 1931 (290), and from 168 to 250 in 1935 (median: 209), respectively.

Among the smallest English-language periodicals in 1931, we found the annual Hongkong University Union Magazine (31.5), the monthly China Journal in Shanghai, and three medical or scientific journals based in Hongkong - Caduceus and Hongkong Naturalist - and Shanghai - China Medical Journal:

In 1935, the list included mostly monthly journals, such as the China Highways, The Shanghailander, and Cathay Magazine (all measuring 25 sq.inches). We also found technical publications such as the Chinese Economic Journal 中國經濟月刊 (Zhongguo jingji yuekan) published by the Bureau of Foreign Trade, which suggests that bureaucrats, scientists and businessmen favored portable formats they could take with them during their professional trips:

The largest English-periodicals consisted essentially in dailies and weeklies published in Shanghai and Hongkong.

The Shanghai Times increased by 50 and square inches between 1931 and 1935, whereas the China Mail, the Hongkong Sunday Herald and the Hongkong Telegraph were reduced by 16, 16 and 10 square inches, respectively:

The largest Japanese newspaper was the Shanghai Mainichi Shimbun 上海每日新聞社 (Shanghai Meiri xinwenshe), which, however, lost 5 square inches between 1931 and 1935. The smallest were the Shanghai Nichi Nichi Shimbun and the Seito Shimpo in Qindbao. Except for the Shanghai Mainichi Shimbun, the size of Japanese newspapers did not change between the two editions.

The largest Russian newspaper was the Harbin Zaria in 1931 (344 sq.inc.) and the Nasha Zaria published in Tianjin in 1935. The smallest was the Harbin-based Russian Kung Pao (275) in 1931, and the Shanghai-based Slovo and Shanghai Zaria in Shanghai in 1935 (both measuring 290 sq. inches).

The smallest German newspaper was the Tianjin-based Deutsch-Chinesische Nachrichten whose size even decreased (from 200 to 168) between 1931 and 1935. The largest was the Deutsche Shanghai Zeitung published in Shanghai.

City

Were compact formats more likely to develop in large cities with modern transportation system and mobile readers? In order to test the hypothesis, we performed a linear regression analysis, setting the average surface of the page as the outcome variable (y) and the city of publication as the explanatory variable (x), setting “Shanghai” as the reference city.

The difference is quite significant in the two directories (p value: 2.474e-06 in 1931, < 2.2e-16 in 1935). The development of portable periodicals was seemingly related to the importance of the city and the social, economic profile of its population. Except for a handful of smaller towns in interior provinces (such as Baotou, Chaoxian, Nantongzhou, Pinghu, Xiaoshan, which moreover, were not statistically significant), periodicals in Shanghai were generally smaller than in those published in other cities.

Size of periodicals and place of publication in 1931:

## 
## Call:
## lm(formula = Page_surface ~ City_py, data = crow_format31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -199.29  -68.01    0.00   56.48  260.92 
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        127.20791    9.43906  13.477  < 2e-16 ***
## City_pyAndong      152.79209   93.91746   1.627 0.104831    
## City_pyAnqing      129.66709   54.76824   2.368 0.018551 *  
## City_pyAomen       139.91709   66.74423   2.096 0.036906 *  
## City_pyBeijing      78.20876   23.96191   3.264 0.001228 ** 
## City_pyChangchun   143.54209   93.91746   1.528 0.127488    
## City_pyChangsha    142.79209   93.91746   1.520 0.129482    
## City_pyChangshu    157.79209   93.91746   1.680 0.093994 .  
## City_pyChangzhou   125.60459   66.74423   1.882 0.060837 .  
## City_pyChengdu      50.76084   47.66492   1.065 0.287769    
## City_pyChongqing    66.57855   28.57817   2.330 0.020497 *  
## City_pyDalian      164.20876   39.29794   4.179 3.87e-05 ***
## City_pyFoshan      162.79209   93.91746   1.733 0.084078 .  
## City_pyFuzhou      112.33376   39.29794   2.859 0.004559 ** 
## City_pyGaoyou       48.57334   66.74423   0.728 0.467342    
## City_pyGuangzhou   148.21006   25.19540   5.882 1.10e-08 ***
## City_pyGuiyang     132.79209   93.91746   1.414 0.158439    
## City_pyHangzhou     79.01865   34.35870   2.300 0.022157 *  
## City_pyHarbin      108.10459   28.57817   3.783 0.000188 ***
## City_pyHeihe       134.47959   93.91746   1.432 0.153234    
## City_pyHengyang    147.79209   66.74423   2.214 0.027572 *  
## City_pyHuzhou      122.79209   93.91746   1.307 0.192079    
## City_pyJiangmen    150.66709   93.91746   1.604 0.109729    
## City_pyJiaxing     142.79209   93.91746   1.520 0.129482    
## City_pyJilin       162.79209   93.91746   1.733 0.084078 .  
## City_pyJinan       135.22066   36.55732   3.699 0.000258 ***
## City_pyJiujiang    152.79209   93.91746   1.627 0.104831    
## City_pyKaifeng     152.79209   93.91746   1.627 0.104831    
## City_pyKunming     129.39923   36.55732   3.540 0.000465 ***
## City_pyLanzhou      77.54209   93.91746   0.826 0.409675    
## City_pyLiuzhou      94.79209   93.91746   1.009 0.313652    
## City_pyNanchang    152.79209   54.76824   2.790 0.005618 ** 
## City_pyNanjing      14.48527   29.71294   0.488 0.626261    
## City_pyNanning      19.04209   66.74423   0.285 0.775615    
## City_pyNantongzhou  -0.02041   93.91746   0.000 0.999827    
## City_pyNingbo      147.79209   66.74423   2.214 0.027572 *  
## City_pyPengpu       72.41709   66.74423   1.085 0.278810    
## City_pyPinghu       -4.64541   93.91746  -0.049 0.960584    
## City_pyQingdao     162.79209   42.84127   3.800 0.000176 ***
## City_pyShantou     109.09209   42.84127   2.546 0.011391 *  
## City_pyShaoxing    145.29209   66.74423   2.177 0.030285 *  
## City_pyShenyang     45.89626   28.57817   1.606 0.109346    
## City_pySuzhou      139.16709   42.84127   3.248 0.001294 ** 
## City_pyTaixian     -13.45791   93.91746  -0.143 0.886155    
## City_pyTaiyuan     165.29209   66.74423   2.477 0.013828 *  
## City_pyTianjin      93.52505   22.04489   4.242 2.96e-05 ***
## City_pyWenzhou     150.29209   66.74423   2.252 0.025073 *  
## City_pyWuchang       3.29209   93.91746   0.035 0.972061    
## City_pyWuhu        160.29209   54.76824   2.927 0.003692 ** 
## City_pyWuxi        137.79209   66.74423   2.064 0.039846 *  
## City_pyWuzhou      147.79209   66.74423   2.214 0.027572 *  
## City_pyXi'an       137.79209   93.91746   1.467 0.143397    
## City_pyXiamen      141.99209   42.84127   3.314 0.001033 ** 
## City_pyXianggang    82.45057   20.02327   4.118 4.97e-05 ***
## City_pyXiaoshan     -6.95791   93.91746  -0.074 0.940993    
## City_pyYangzhou    122.26084   66.74423   1.832 0.067992 .  
## City_pyYingkou     113.29209   66.74423   1.697 0.090674 .  
## City_pyZhenhai     138.79209   66.74423   2.079 0.038438 *  
## City_pyZhenjiang   121.29209   42.84127   2.831 0.004956 ** 
## City_pyZhifu       130.37543   54.76824   2.380 0.017924 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 93.44 on 295 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.3161, Adjusted R-squared:  0.1793 
## F-statistic: 2.311 on 59 and 295 DF,  p-value: 2.474e-06

Size of periodicals and place of publication in 1935:

## 
## Call:
## lm(formula = Page_surface ~ City_py, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -219.13  -49.23    0.00   44.76  271.89 
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        116.2341     6.2654  18.552  < 2e-16 ***
## City_pyAndong      163.7659    85.9066   1.906 0.057100 .  
## City_pyAnqing      149.5409    38.8252   3.852 0.000130 ***
## City_pyAomen       155.0159    43.2947   3.580 0.000372 ***
## City_pyBaotou       -0.2966    85.9066  -0.003 0.997246    
## City_pyBeijing     100.5209    18.2451   5.509 5.41e-08 ***
## City_pyChangchun   154.5159    85.9066   1.799 0.072592 .  
## City_pyChangde      92.8492    49.8613   1.862 0.063086 .  
## City_pyChangsha    141.9534    35.5345   3.995 7.31e-05 ***
## City_pyChangshu    145.4325    49.8613   2.917 0.003673 ** 
## City_pyChangzhou   116.2659    35.5345   3.272 0.001131 ** 
## City_pyChaoxian     -3.7341    85.9066  -0.043 0.965344    
## City_pyChengdu      69.3128    30.9328   2.241 0.025419 *  
## City_pyChongde     149.7659    85.9066   1.743 0.081799 .  
## City_pyChongming     4.1409    60.9065   0.068 0.945819    
## City_pyChongqing    85.4971    20.1566   4.242 2.58e-05 ***
## City_pyDalian      175.1825    35.5345   4.930 1.07e-06 ***
## City_pyDanyang      13.7659    85.9066   0.160 0.872746    
## City_pyDatong        7.5159    85.9066   0.087 0.930313    
## City_pyFenghua     124.2659    85.9066   1.447 0.148569    
## City_pyFengxian      0.7659    85.9066   0.009 0.992890    
## City_pyFoshan      173.7659    85.9066   2.023 0.043558 *  
## City_pyFuzhou      106.7103    29.2385   3.650 0.000286 ***
## City_pyGanzhou     161.6409    85.9066   1.882 0.060391 .  
## City_pyGaoyou        3.4065    43.2947   0.079 0.937313    
## City_pyGuangzhou   179.0203    19.7183   9.079  < 2e-16 ***
## City_pyGuisui      157.8909    60.9065   2.592 0.009772 ** 
## City_pyGuiyang     138.6409    60.9065   2.276 0.023191 *  
## City_pyHaikou      149.7659    85.9066   1.743 0.081799 .  
## City_pyHaimen      135.3909    38.8252   3.487 0.000525 ***
## City_pyHangzhou     95.5207    24.5749   3.887 0.000113 ***
## City_pyHarbin      117.7159    38.8252   3.032 0.002538 ** 
## City_pyHechuan      -5.7341    85.9066  -0.067 0.946805    
## City_pyHefei        -6.2654    60.9065  -0.103 0.918103    
## City_pyHeihe       145.4534    85.9066   1.693 0.090961 .  
## City_pyHengyang    139.6721    60.9065   2.293 0.022191 *  
## City_pyHexian        9.7659    85.9066   0.114 0.909531    
## City_pyHuangyan      3.0159    85.9066   0.035 0.972007    
## City_pyHuizhou     125.2659    85.9066   1.458 0.145335    
## City_pyHuzhou      125.7659    60.9065   2.065 0.039375 *  
## City_pyJiangmen    155.7034    60.9065   2.556 0.010828 *  
## City_pyJiangyin    124.4325    49.8613   2.496 0.012851 *  
## City_pyJiaxing     154.5159    60.9065   2.537 0.011443 *  
## City_pyJilin       173.7659    85.9066   2.023 0.043558 *  
## City_pyJinan       110.8016    23.7400   4.667 3.79e-06 ***
## City_pyJingjiang     9.7659    85.9066   0.114 0.909531    
## City_pyJinhua      133.5159    85.9066   1.554 0.120681    
## City_pyJiujiang     98.3492    49.8613   1.972 0.049031 *  
## City_pyKaifeng     112.0784    49.8613   2.248 0.024962 *  
## City_pyKaiping     134.0159    85.9066   1.560 0.119299    
## City_pyKunming     140.1750    26.5818   5.273 1.89e-07 ***
## City_pyKunshan     -12.2341    85.9066  -0.142 0.886804    
## City_pyLanxi       117.7659    85.9066   1.371 0.170947    
## City_pyLanzhou     126.1409    60.9065   2.071 0.038794 *  
## City_pyLiling        8.6409    85.9066   0.101 0.919915    
## City_pyLishui      163.7659    85.9066   1.906 0.057100 .  
## City_pyLiuzhou     105.7659    85.9066   1.231 0.218755    
## City_pyLongyou       8.6409    85.9066   0.101 0.919915    
## City_pyLuoyang     126.7659    85.9066   1.476 0.140586    
## City_pyNanchang    163.0992    35.5345   4.590 5.44e-06 ***
## City_pyNanjing      50.3214    17.6389   2.853 0.004487 ** 
## City_pyNanning      30.0159    60.9065   0.493 0.622326    
## City_pyNanping      -5.7341    85.9066  -0.067 0.946805    
## City_pyNantongzhou  64.4891    32.9837   1.955 0.051040 .  
## City_pyNingbo       92.2659    38.8252   2.376 0.017803 *  
## City_pyPengpu      117.3284    43.2947   2.710 0.006927 ** 
## City_pyPinghu      117.7659    85.9066   1.371 0.170947    
## City_pyQingdao     145.3100    21.7039   6.695 5.08e-11 ***
## City_pyQingpu        9.7659    85.9066   0.114 0.909531    
## City_pyQingyuan    183.7659    85.9066   2.139 0.032840 *  
## City_pyQuanzhou    149.7659    85.9066   1.743 0.081799 .  
## City_pyRugao        22.7034    85.9066   0.264 0.791658    
## City_pyShantou     144.0575    35.5345   4.054 5.72e-05 ***
## City_pyShaoxing    156.2659    60.9065   2.566 0.010547 *  
## City_pyShashi      100.5575    49.8613   2.017 0.044181 *  
## City_pyShengxian     9.7659    85.9066   0.114 0.909531    
## City_pyShenyang    108.3730    32.9837   3.286 0.001079 ** 
## City_pyShunyi        9.7659    85.9066   0.114 0.909531    
## City_pySiyang       -1.4841    85.9066  -0.017 0.986222    
## City_pySongjiang    28.7659    85.9066   0.335 0.737858    
## City_pySuining     -41.2341    85.9066  -0.480 0.631416    
## City_pySuzhou      152.4266    32.9837   4.621 4.70e-06 ***
## City_pyTaicang     134.0159    85.9066   1.560 0.119299    
## City_pyTaishan     170.1409    60.9065   2.793 0.005386 ** 
## City_pyTaixian      13.1409    60.9065   0.216 0.829254    
## City_pyTaiyuan      69.4048    29.2385   2.374 0.017932 *  
## City_pyTangshan      6.3284    85.9066   0.074 0.941302    
## City_pyTaoyuan      13.2659    85.9066   0.154 0.877330    
## City_pyTianjin      47.7480    13.7501   3.473 0.000554 ***
## City_pyWanxian     117.7659    85.9066   1.371 0.170947    
## City_pyWenzhou     141.8492    35.5345   3.992 7.39e-05 ***
## City_pyWuchang      12.0159    85.9066   0.140 0.888810    
## City_pyWuhu        166.5159    49.8613   3.340 0.000893 ***
## City_pyWujiang       9.7659    85.9066   0.114 0.909531    
## City_pyWuxi         86.2242    35.5345   2.426 0.015548 *  
## City_pyWuzhou      155.7659    49.8613   3.124 0.001873 ** 
## City_pyXi'an       149.5159    43.2947   3.453 0.000594 ***
## City_pyXiamen      119.0020    29.2385   4.070 5.35e-05 ***
## City_pyXianggang   128.3795    16.1771   7.936 1.08e-14 ***
## City_pyXiangtan    117.7659    85.9066   1.371 0.170947    
## City_pyXiangyang    -2.4841    85.9066  -0.029 0.976941    
## City_pyXiaoshan      4.0159    85.9066   0.047 0.962731    
## City_pyXinghua      20.2659    85.9066   0.236 0.813589    
## City_pyXingning    149.7659    85.9066   1.743 0.081799 .  
## City_pyXinning       5.2659    85.9066   0.061 0.951143    
## City_pyXuancheng    88.6721    60.9065   1.456 0.145967    
## City_pyXuzhou      101.5992    49.8613   2.038 0.042038 *  
## City_pyYancheng     67.0159    60.9065   1.100 0.271653    
## City_pyYangzhou    132.7659    30.9328   4.292 2.07e-05 ***
## City_pyYichang     140.9742    49.8613   2.827 0.004855 ** 
## City_pyYuyao       130.7659    85.9066   1.522 0.128505    
## City_pyZhangzhou   117.7659    85.9066   1.371 0.170947    
## City_pyZhengzhou    97.3492    49.8613   1.952 0.051370 .  
## City_pyZhenhai     145.0159    60.9065   2.381 0.017588 *  
## City_pyZhenjiang   119.6284    22.9920   5.203 2.72e-07 ***
## City_pyZhifu       157.2034    35.5345   4.424 1.16e-05 ***
## City_pyZixing       -1.4841    85.9066  -0.017 0.986222    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 85.68 on 582 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.379,  Adjusted R-squared:  0.2552 
## F-statistic: 3.062 on 116 and 582 DF,  p-value: < 2.2e-16

In order to refine our model, we classified the cities using the following categories:

Geographic location:

Geographic Location
Code Description
CT Central
NNE North-Northeast
NW Northwest
SE Southeast (including Hongkong/Macao)
SW Southwest


City profile:

City profile
Code Description
0 Treaty port and metropolis of international dimension
1 Other large urban center, provincial or national capital
2 Smaller or interior town

The cities were coded as followed:

library(readr)
city_code <- read_delim("city_code.csv",
                        ";", escape_double = FALSE, trim_ws = TRUE)
## 
## ── Column specification ────────────────────────────────────────────────────────
## cols(
##   City_zh = col_character(),
##   Province = col_character(),
##   City_py = col_character(),
##   Profile = col_double(),
##   Location = col_character()
## )
crow_page_city <- crowdata %>% select(Year, Title_eng, Title_zh, City_zh, City_py, Province, Page_surface)
code <- city_code %>% select(City_zh, Profile, Location)

page_city <- right_join(crow_page_city, code)
## Joining, by = "City_zh"
page_city <- page_city %>% mutate(Profile = as.factor(Profile))
page_city$Profile <- relevel(page_city$Profile,ref="0")
page_city <- page_city %>% mutate(Location = as.factor(Location))
page_city$Location <- relevel(page_city$Location,ref="SE")

page_city31 <- page_city %>% filter(Year == "1931")
page_city35 <- page_city %>% filter(Year == "1935")

page_city %>% 
  arrange(Year, Profile)

Based on this classificaiton, we perform a linear regression analyses again using the two new predicting variables separately and using an additive or interaction model. We set “0” as the reference profile, and “SE” as the reference location.

The difference based on the city profile was indeed statistically significant, especially in 1935 (p value: 0.0002762 in 1931, 1.917e-08 in 1935). Periodicals published in treaty ports and metropolises with a strong international profile were generally more portable than periodicals published elsewhere in China. Those published in non treaty port but major cities were 26 and 45 sq. inches larger in 1931 and 1935, respectively. Those published in smaller interior towns were 54 sq. inches larger in 1931, but the gap had decreased by 1935 (43 sq.inches in 1935), which suggests that smaller towns tended to aligned on modern standards at a faster pace than other urban centers.

Size of periodicals and city profile in 1931:

## 
## Call:
## lm(formula = Page_surface ~ Profile, data = page_city31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -202.41  -93.01   30.84   84.96  277.21 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  185.044      6.895  26.839  < 2e-16 ***
## Profile1      26.970     13.741   1.963   0.0505 .  
## Profile2      54.863     13.814   3.972 8.65e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 100.9 on 354 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.04524,    Adjusted R-squared:  0.03985 
## F-statistic: 8.387 on 2 and 354 DF,  p-value: 0.0002762

Size of periodicals and city profile in 1935:

## 
## Call:
## lm(formula = Page_surface ~ Profile, data = page_city35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -188.43  -94.20   26.38   73.57  217.23 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  170.891      5.077  33.662  < 2e-16 ***
## Profile1      45.536      9.535   4.776 2.19e-06 ***
## Profile2      43.230      8.639   5.004 7.12e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 96.86 on 697 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.04971,    Adjusted R-squared:  0.04698 
## F-statistic: 18.23 on 2 and 697 DF,  p-value: 1.917e-08

Similarly, there was a statistically significant difference in size depending on the geographic situation, though not as important as the city profile (p value: 0.0009293 in 1931, 0.001328 in 1935). Periodicals published in Southeastern cities were clearly more compact than those published in other regions.

Size of periodicals and geographic situation in 1931:

## 
## Call:
## lm(formula = Page_surface ~ Location, data = page_city31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -193.38  -94.01   41.37   80.43  277.68 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  184.572      6.767  27.277  < 2e-16 ***
## LocationCT    88.741     32.592   2.723 0.006796 ** 
## LocationNNE   44.809     12.406   3.612 0.000348 ***
## LocationNW    50.303     71.610   0.702 0.482857    
## LocationSW    27.775     19.907   1.395 0.163818    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 100.8 on 352 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.05141,    Adjusted R-squared:  0.04063 
## F-statistic: 4.769 on 4 and 352 DF,  p-value: 0.0009293

The stronger difference in 1931 referred to central (88 sq. inches more), then northwestern (+50), northeastern (+44) and finally, southwestern provinces (+27). Interestingly, rural provinces in the Southwest were closer to treaty port standards than one would expect. Further research is needed to unveil the processes underlying these patterns of similarity.

Size of periodicals and geographic situation in 1935:

## 
## Call:
## lm(formula = Page_surface ~ Location, data = page_city35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -173.56  -90.26   31.77   85.20  207.33 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  180.795      4.631  39.041  < 2e-16 ***
## LocationCT    36.537     15.517   2.355  0.01882 *  
## LocationNNE   28.386      9.332   3.042  0.00244 ** 
## LocationNW    64.975     33.071   1.965  0.04984 *  
## LocationSW    31.190     14.644   2.130  0.03353 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 98.24 on 695 degrees of freedom
##   (3 observations deleted due to missingness)
## Multiple R-squared:  0.0253, Adjusted R-squared:  0.01969 
## F-statistic:  4.51 on 4 and 695 DF,  p-value: 0.001328

By 1935, the gap with northwestern and southwestern provinces had deepened (+64 and + 31 sq. inches, respectively), whereas central (+36) and most importantly, northeastern periodicals (+28) had become closer to the standards established in Southeastern cities. Compact periodicals developed at a faster rate in Northeastern China than anywhere else.

In addition, there were important variations in size across and within cities.

The interactive graphs below are enriched with statistics for each language category. You can zoom in and compare the figures across cities and between the two directories. The boxplots visualize the range of formats in each city, whereas the barplots shows the distribution between the three main categories:

Distribution of format in each city:

Format categories, by city:

Page dimensions

As the scatterplots reveal, most periodicals presented an elongated format. The vertical dimension (y) usually exceeded the width (x) (dots above the line):

A handful of periodicals, however, adopted a horizontal format. Their width was superior to their length (dots below the line). The few dots that are very close to the line refer to periodicals with an almost square format (their width almost equaling their length). The Chin Chin Movie Monthly in 1935 even had a perfectly square format (see the dot located exactly on the line).

The interactive plots below are enriched with detailed statistics for each dot:

If we take into account the frequency of publication:


Taking language into account:


The scatterplots show no clear relation between the dimensions of the page, the frequency and the language of publication. Among vertical periodicals, dailies usually presented the most elongated pages, whereas the least frequent periodicals (monthlies, semi-monthlies, quarterlies) appeared shorter, with a corresponding smaller width. Weekly and intermediate frequency stood in between.

Horizontal pages (ratio < 1) referred to a variety of periodicals, including the German daily Deutsche Shanghai Zeitung, the English medical quarterly Caduceus and a variety of Chinese periodicals wity varying frequency.

Periodicals with horizontal pages (ratio < 1):

Among the few publications with a squared or almost squared format, we found three periodicals devoted to cinema, including the illustrated magazine Chin Chin Movie Monthly (Chin Chin Dien Yin Yu Kan) 青青電影畫報月刊 (Qingqing dianying huabao yuekan) and three tabloids - Radio Film Daily News (Dien Yin Jih Pao) 電影日報 (Dianying ribao), Radio Movie Daily News (Dien Shun Jih Pao) 電聲日報 (Diansheng ribao) and Camera News (Ka Ma La) 開麥拉 (Kaimaila) - all established between 1932 and 1934. Square-page movie magazines appeared to be specific to Shanghai, which was the center of the emerging movie industry in the 1930s (Kerlan).

The two periodicals with perfectly square pages:

Page and columns

How did the size of pages relate to the number of columns they displayed?

The scatterplots below correlate the number of columns (y) and the surface of the page (x), measured in square inches. It seems that the number of columns increased with the surface of the page:

The interactive plots below are enriched with detailed statistics for each dot:

In order to modelize the relation, we run a simple linear regression analysis (lm), setting the surface of the page as the outcome variable (y) and the number of columns as the explanatory variable (x).

The relation was statistically significant and positive in the two directories (p value: < 2.2e-16 in 1931, 1.459e-11 in 1935), which means that columns did multiply as the page expanded. More precisely, when the page increased by 20-21 square inches, it gained one additional column.

Page surface and number of columns in 1931:

## 
## Call:
## lm(formula = Page_surface ~ Col_nbr_avrg, data = crow_format31)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -144.246  -47.466   -4.246   47.900  146.025 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept)    73.513     12.056   6.098  3.4e-09 ***
## Col_nbr_avrg   21.073      1.515  13.907  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 64.61 on 292 degrees of freedom
##   (64 observations deleted due to missingness)
## Multiple R-squared:  0.3985, Adjusted R-squared:  0.3964 
## F-statistic: 193.4 on 1 and 292 DF,  p-value: < 2.2e-16

Page surface and number of columns in 1935:

## 
## Call:
## lm(formula = Page_surface ~ Col_nbr_avrg, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -125.25  -71.17  -10.22   64.22  146.34 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   107.546     19.748   5.446 5.95e-07 ***
## Col_nbr_avrg   20.529      2.593   7.919 1.46e-11 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 76.27 on 77 degrees of freedom
##   (623 observations deleted due to missingness)
## Multiple R-squared:  0.4488, Adjusted R-squared:  0.4417 
## F-statistic: 62.71 on 1 and 77 DF,  p-value: 1.459e-11

In the next section, we will examine more closely the layout of periodicals, defined by the number and the dimensions of columns displayed on their pages.

Layout

In the last edition of the directory (1937), Crow explained that:

The column of a Chinese paper is worked out by the height of the line, as Chinese types run from top to bottom vertical. Big Shanghai papers have their standardised columns at 5 inches for main editions and 2 1/2 inches for local edition supplements. Each line in a column has 40 Chinese characters of the No.5 type, or 8 characters make a depth of one inch. All advertisements in the main editions of the larger papers must be five inches deep, or a multiple of five inches. This requirement need not to be observed in the case of the smaller provincial papers which will accommodate almost any size of advertisement. The most common sizes are 6X9 inches (quarter page) or 4X6 inches (one eighth page).

(Carl Crow, Inc. “Newspaper Mechanical Requirements.”, p.1)

How did the pages of periodicals look like? How many columns, of what size, did they include? How did the layout of evolve over time? Did it become more standardized, or more diversified?

Number of columns

The histograms reveal significant discrepancies between periodicals. The number of pages ranges from 1 to 13 in 1931, from 2 to 13 in 1935, with a slightly decreasing average from 8 in 1931 to 7 in 1935. The histograms suggest a reduction in the number of columns. While periodicals with eight to ten columns were the most frequent in 1931, those with five to ten columns clearly dominated in 1935. The was also a remarkable growth of two-column periodicals. Although they did not really increased in number, they proportionally eclipsed their multicolumn counterparts.

Given the scarcity of information, however, these figures should be taken with caution. The information is missing for 63 periodicals in 1931 (58%) and 622 in 1935 (94%). The huge gap between the two editions prevents us from drawing sound conclusions about changes in layout over time.

Detailed statistics in 1931:

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##       1       6       8       8      10      13      63
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##       1       6       7       7       9      12      58

Detailed statistics in 1935:

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##       2       4       7       7       8      13     622
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##       0       1       1       1       1       2      94

Based on the histograms, we devised five categories to facilitate further analyses:

The table confirms the growth of periodicals with fewer columns (less than 3) at the expense of those with more than eight columns (from 31% to 21%). Two-column periodicals grew from 7 to 20%. Periodicals with three to five columns stagnated around 11%. Those between five and eight columns maintained a short those declining majority (from 51% to 48%).

Japanese dailies in Manchuria (Manchu Nippon, Dairen Shimbun) and Shanghai (Shanghai Mainichi Shimbun, Shanghai Nichi Nichi Shimbun, Shanghai Nippo Sha) featured the largest number of columns (13). The Japanese model seems to influence Chinese dailies in Japanese-occupied Manchuria, such as North Eastern Public Press 東北民衆報 (Dongbei minzhong bao) in Shenyang and Manchu News 滿洲報 (Manzhoubao) in Dalian, which displayed twelve columns. The list also includes a handful of Chinese dailies, mostly in Northern cities (Qingdao, Tianjin, Harbin), and the Hongkong-based weekly South China Sunday News 南華星期報 (Nanhua xingqibao).

While Japanese periodicals still topped the list in 1935, an increasing number of Chinese dailies based in Shanghai adopted the 12-column layout:

In the two directories, the periodicals with fewer columns were mostly English-language weeklies. The only periodical with a single column in 1931 was the English medical quarterly The Caduceus based in Hongkong. The majority of two-page periodicals consists in English-language weeklies based in Shanghai, Guangzhou and Hongkong. The only exception in 1931 was the Chinese daily Strong Group Press 羣强報 (Qunqiangbao) published in Beijing.

Periodicals with less than three columns in 1931 :

Periodicals with less than three columns in 1935 :

These observations suggest some relation between the layout, the language and the frequency of publication. Periodicals with shorter periodicity (especially daily newspapers) seem to feature a larger number of column than less frequent publications. Moreover, Japanese multicolumn newspapers contrasted with English weeklies with very few columns.

To what extent did the periodicity and the language of periodicals impact on their layout? The next two sections aim to analyze in a more systematic way this seemingly triangular relation.

Periodicity

How did the frequency of publication impact on the layout of periodicals?

In the two directories, dailies generally contained more columns than less frequent periodicals. This largely reflects the proportionate size of the page. As we have seen earlier, daily newspapers generally contained larger pages than less frequent periodicals, which were more inclined to adopt compact formats.

In addition, dailies showed greater greater variability in their layout than other periodicals. In 1931, the number of columns in daily newspapers could vary from 2 to 13 pages, compared to 2 to 12 for weeklies and 2 to 4 for monthlies. This largely reflect the variety of language, as we will see later. Moreover, variability tended to decrease over time. In 1935, the possible number of columns in dailies ranged from 4 to 13, from 2 to 8 for weeklies. It remained stable for monthlies. Notwithstanding the gap in information, this confirms the standardization process which we highlighted earlier.

The interactive boxplots are enriched with statistics for each category of periodicals. You can zoom in and compare the figures across periodicals and between the two directories:

In order to systematize the analysis, we applied a linear regression model (lm) with the number of columns as the outcome variable (y) and periodicity as the explanatory variable, setting daily newspapers as the reference category.

We observe a strong correlation in the two directories (p value <2e-16 in 1931, 1.358e-08). The layout was clearly a factor of differentiation between periodicals with different periodicity. Daily newspapers clearly contained a larger number of columns than less frequently published periodicals.

Correlation between the number of columns and the frequency of publication in 1931:

## 
## Call:
## lm(formula = Col_nbr_avrg ~ Periodicity, data = crow_format31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -6.1423 -2.1423 -0.1423  1.8577  6.8043 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            8.1423     0.1391  58.548  < 2e-16 ***
## PeriodicityAnnual     -6.1423     2.1545  -2.851  0.00467 ** 
## PeriodicityBiweekly   -1.4756     1.2490  -1.181  0.23843    
## PeriodicityMonthly    -5.4756     1.2490  -4.384 1.64e-05 ***
## PeriodicityQuarterly  -7.1423     2.1545  -3.315  0.00103 ** 
## PeriodicityTabloid    -1.1423     1.5266  -0.748  0.45493    
## PeriodicityWeekly     -2.9466     0.3462  -8.512 9.54e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 2.15 on 288 degrees of freedom
##   (63 observations deleted due to missingness)
## Multiple R-squared:  0.2678, Adjusted R-squared:  0.2526 
## F-statistic: 17.56 on 6 and 288 DF,  p-value: < 2.2e-16

In 1931, all categories featured fewer columns on their page than daily newspapers. The difference was particularly huge with quarterlies (- 7 columns), annuals (- 6), monthlies (- 5) and weeklies (-2.9) The difference was less significant with more frequent periodicals, namely tabloids (-1.1) and biweeklies (-1.4). Note: The figures for biweeklies and tabloids, however, should be taken with caution, given that the p-value for these categories is superior to 0.05.

Correlation between the number of columns and the frequency of publication in 1935:

## 
## Call:
## lm(formula = Col_nbr_avrg ~ Periodicity, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.1379 -1.6667 -0.6667  1.5155  4.8621 
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           8.1379     0.3402  23.922  < 2e-16 ***
## PeriodicityAnnual    -5.1379     2.6131  -1.966   0.0530 .  
## PeriodicityBiweekly  -6.1379     2.6131  -2.349   0.0215 *  
## PeriodicityMonthly   -5.5379     1.2076  -4.586 1.77e-05 ***
## PeriodicityWeekly    -4.4713     0.7505  -5.958 7.72e-08 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 2.591 on 75 degrees of freedom
##   (622 observations deleted due to missingness)
## Multiple R-squared:  0.428,  Adjusted R-squared:  0.3975 
## F-statistic: 14.03 on 4 and 75 DF,  p-value: 1.358e-08

The gap persisted in 1935 but the relation was not as linear as in 1931. The stronger difference was now between dailies and biweeklies, which contained six columns less. Monhtlies and annuals included five columns less and weeklies, four. The lack of information in this edition, however, may disturb the relation we observed in the first directory. Overall, there was a clear, almost perfectly linear relation between the number of pages and the frequency of publication. The more frequent the periodical, the larger the number of columns it contained.

How did language impact on the layout of periodicals?

Language

The histograms suggest that Japanese newspapers featured the larger number of columns, whereas English periodicals contained fewer columns. Chinese periodicals presented the largest spectrum. Other foreign languages stood in between.

Moreover, while Japanese newspapers held the record number of columns, English periodicals displayed the greatest variability in their layout.

The interactive boxplots are enriched with statistics for each category of periodicals. You can zoom in and compare the figures across periodicals and between the two directories:

The number of columns in English periodicals may vary from one to eight in 1931, between two and eight in 1935. Due to their statistical importance, Chinese periodicals came second (from two to twelve, four to twelve), though their variability decreased between the two editions.

In 1931, Russian newspapers included between five - Russian Kung Pao (Harbin) - to eight columns - The Zaria (Harbin). German newspapers in 1935 contained between four - Deutsche Shanghai Zeitung (Shanghai) - and six columns - Deutsch-Chinesische Nachvichten (Tianjin). The French Journal de Shanghai 法文上海日報 (Fawen Shanghai ribao) contained eight pages.

In order to modelize the relation, we performed a linear regression analysis setting the number of columns as the response variable (y) and language as the predictor variable (x), setting “Chinese” as the reference language category.

The difference was highly significant in the two directories (p value: 1.055e-15 in 1931, 2.617e-09 in 1935). The layout was clearly a factor of differentiation between languages.

Number of columns and language in 1931:

## 
## Call:
## lm(formula = Col_nbr_avrg ~ Language, data = crow_format31)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -5.8293 -1.8293  0.1707  2.1707  4.1707 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        7.8293     0.1395  56.139  < 2e-16 ***
## LanguageEnglish   -2.9150     0.3952  -7.377 1.73e-12 ***
## LanguageFrench     0.1707     2.1918   0.078    0.938    
## LanguageGerman    -1.8293     2.1918  -0.835    0.405    
## LanguageJapanese   5.0041     0.9038   5.537 6.93e-08 ***
## LanguageRussian   -0.9959     0.9038  -1.102    0.271    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 2.187 on 289 degrees of freedom
##   (63 observations deleted due to missingness)
## Multiple R-squared:  0.2395, Adjusted R-squared:  0.2263 
## F-statistic:  18.2 on 5 and 289 DF,  p-value: 1.055e-15

Number of columns and language in 1935:

## 
## Call:
## lm(formula = Col_nbr_avrg ~ Language, data = crow_format35)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.0645 -2.0645 -0.0323  2.1622  3.9355 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       8.06452    0.44864  17.975  < 2e-16 ***
## LanguageEnglish  -3.22668    0.60821  -5.305 1.13e-06 ***
## LanguageFrench   -0.06452    2.53789  -0.025   0.9798    
## LanguageGerman   -3.06452    1.82238  -1.682   0.0969 .  
## LanguageJapanese  4.76882    1.11410   4.280 5.52e-05 ***
## LanguageRussian  -1.06452    1.51035  -0.705   0.4831    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 2.498 on 74 degrees of freedom
##   (622 observations deleted due to missingness)
## Multiple R-squared:  0.4754, Adjusted R-squared:   0.44 
## F-statistic: 13.41 on 5 and 74 DF,  p-value: 2.617e-09

To sum up, Chinese periodicals generally contained more columns than their foreign counterparts, except for Japanese newspapers that featured 5 columns more in 1931, 4.7 in 1935. The difference with French periodicals remained tiny (between 0.17 and -0.6 in 1935), which suggests that French newspapers included more columns than other languages. The French press in China, however, was not statistically important (p value: 0.938 in 1931, 0.9798 in 1935). The strongest negative difference was with English (-2.9 in 1931, -3.22 in 1935), and less significantly, German periodicals (-1.8 in 1931, -3 in 1935).

Variability

The layout was usually stable from one issue to the other. The number of column varied in nine periodicals, all Chinese-language daily newspapers:

The scatter plots below visualize in greater detail the deviation between the minimum and the maximum number of columns:

The interactive plots below are enriched with detailed statistics for each dot:

The number of columns doubled in three periodicals, from three to six - Moukden Times 盛京時報 (Shengjing shibao), Red Sun News 旭日報 (Xuribao) - or four to eight - Kaoyu Daily Press 高郵日報 (Gaoyou ribao). Three others gained two columns, increasing from four to six - Fearless Press 大無畏報 (Dawu weibao), Social News 社會新報 (Shehui xinbao) - or six to eight - New National Prosperity News 新國華報 (Xinguo huabao), an illustrated newspaper in Guangzhou. Two others - Industrial And Commercial Daily Press 工商日報 (Gongshang ribao) in Hongkong and Tientsin Press 庸報 (Yongbao) in Tianjin - may include four more columns (from eight to twelve). The Public Critic Press 公評報 (Gongpingbao) in Guangzhou held the record with ten additional columns (from six to sixteen):

It appeared that Chinese editors were more likely to change the layout than their foreign colleagues:

The dots that deviate from the line all referred to Chinese daily newspapers:

Among periodicals with unstable layout, a significant proportion was published in smaller towns in interior provinces (Gaoyou, Zhenhai, Kunming) where standardization was not necessarily the rule, due to limited access to modern printing technology and shortage of human skills. Six of them, however, were based in major coastal cities (Guangzhou, Hongkong, Shenyang, Tianjin). Variability in layout, therefore, reflected editorial choices as much as technical constraints. Further research is needed to illuminate such important changes in the daily making of newspapers.

Dimensions of columns

As shown on the graphs below, the majority of periodicals featured vertical columns - their length y exceeding their width x (dots above the line) - but some of them displayed horizontal columns (dots above the line).

Average width and length:

The interactive plots below are enriched with detailed statistics for each dot:

Minimum width and length:

Maximum width and length:

In order to better interpret the difference, we computed the ratio x/y (width/height, or horizontal/vertical dimension). The larger the ratio, the more elongated the column.

The histograms reveal huge discrepancies between periodicals, especially in 1931. The ratio ranged from 0.0014 to 75. Most periodicals showed a ratio between 0.5 and 1.

Detailed statistics in 1931 and 1935:

##     Min.  1st Qu.   Median     Mean  3rd Qu.     Max.     NA's 
##   0.0133   3.2240   5.1000  10.0784   7.0000 712.5000       67
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##   0.080   0.600   1.949   4.396   6.750  31.000     626

The ten periodicals with the most elongated columns:

Except for the English monthly Hexagon based in Shanghai, the most elongated columns were to be found in Chinese daily newspapers. Most of them were based in Hongkong, which suggests that newspaper men in the British colony developed a specific model layout for Chinese newspapers in this city. Further research is needed to understand the social and technical reasons for this specificity.

On the opposite, the ten periodicals with horizontally striped pages (ratio < 0.086) include mostly Chinese newspapers based in Nanjing and less prominent urban centers, such as Zhenhai, Xiaoshan or Yingkou:

Periodicity

Among the periodicals with horizontal columns, we found a majority of dailies or tabloids, a few monthlies and one weekly. The majority of periodicals, however, are very close to the line, which means that the difference was barely significant and rendered into to square-like columns. Weeklies displayed columns with the most balanced dimensions. A handful of monthlies and dailies presented a more clearly horizontal layout (bottom right-hand corner).

Language

The scatterplots below suggest a much more significant correlation between the language of periodicals and their layout. German and Japanese periodicals favored the horizontal layout, whereas vertical columns appeared more fitted to Chinese characters (top left hand corner). English periodicals stood in between and were the closest to the line, which suggests their preference for square columns.

Variability

The dimensions of columns varied in thirteen periodicals. In most cases, only the width varied, while the length remained constant. The North-China Daily News was the only periodical to display columns with fixed width but varying length. In its Sunday edition, the North-China Sunday News - both the width and the length varied. From these observations, it appeared that variation in the dimensions of columns was associated with weekly and other special editions.

In order to better examine the periodicals in question, we computed the ratio x min/x max and we select those whose ratio was different from 1:

The list includes only Chinese daily newspapers, except for the English weekly North-China Sunday News:

The language did not influence variation:

The scatterplots suggest that the variability decreased between 1931 and 1935. In 1935, there were fewer dots that deviated from the line. Notwithstanding the gap in information between the two editions, this confirms the standardization process we observed earlier.

Concluding remarks

To conclude this essay, we observed that large-size periodicals with few pages dominated and even progressed in the early 1930s. This reflects the general growth of the daily press in Republican China, as we have shown in a previous essay.

The length of periodicals was mostly determined by their frequency, their content and their readers’ profile. The least frequent and most specialized publications were usually the most substantial. On the opposite, daily newspapers devoted to general or local news - i.e. constantly changing content that had to printed in emergency - included fewer pages. Variation in volume reflected the editors’ propensity to issue supplements and special editions, and the tremendous growth of advertisements as a major source of revenue for newspapers.

The size of periodicals depended on their periodicity and their language. The most frequently published (daily) were the largest, whereas less frequent periodicals were of smaller size. Foreign-language periodicals, especially Japanese, French, Russian and German, were larger than their Chinese and English counterparts. The pages of Chinese newspapers adopted a more elongated shape than British broadsheets. Square-page movie magazines emerged as a specific “genre” in Shanghai.

Although large-size broadsheets remained the majority, compact and book-modeled periodicals experienced a dramatic growth in the early 1930s. Instead of a linear process of miniaturization (Chin, 2014), however, we observed a more complex pattern. The numerical growth of broadsheets did not compensate their decline in proportion, as compact periodicals with intermediate periodicity (bi- or semi-weekly, tabloids) developed at a higher pace during the same period. Moreover, while some periodicals did shrink, others increased in size between the two editions. Further research is needed to determine whether this process of reduction was a response to the shortage of ink and paper and the rising costs of production during the economic depression; whether it expressed editors’ concerns for accommodating changing reading practices and urban dwellers’ preference for more portable formats; or whether it reflected more profound changes in the press industry and the production of newspapers. As we will see in a future essay, we observe a greater proximity between news editors and book publishers in the Republican period. It remains to be established whether such convergence was just of process of imitation or whether it relied on the concrete exchange of materials, personnel and equipment.

The layout also depended on the periodicity and the language. Less frequent publications (annuals, quarterlies monthlies) usually contained fewer columns than dailies and more frequent periodicals. It should be noted than Japanese newspapers displayed more columns than any other languages. The dimensions of columns also varied across languages, which suggests that foreign editors relied on their own imported materials and printing technology to replicate the model layout that was standard in their country of origin. In contrast to the length, the layout of periodicals rarely varied between issues. It was obviously easier to augment or reduce the volume of pages than reshaping their very face. Any change in the number and dimensions of columns would have meant deep changes in technical equipment or editorial choices.

Ultimately, this research reveals that the Chinese press was engaged in a general process of material stabilization and standardization during the Republic (1912-49), as evidenced by the decreasing variability in the length (number of pages), the format (dimensions of pages) and the layout (number and dimensions of columns) of periodicals between 1931 and 1935.

In order to better understand their physical and material characteristics, the next essay will investigate the production and circulation of periodicals.

References

Chin, Sei Jeong. “Print Capitalism, War, and the Remaking of the Mass Media in 1930s China.” Modern China 40, no. 4 (2014): 393–425.

Hong Yu 洪煜. Jindai Shanghai xiaobao Yu shimin wenhua yanjiu 近代上海小报与市民文化研究: 1897-1937 (A Study on Small format Newspapers and Citizens’ Culture in Modern Shanghai). Shanghai: Shanghai shudian chuban she, 2007.

Kerlan-Stephens, Anne, and Christian Delage. Hollywood à Shanghai: l’épopée des studios Lianhua, 1930-1948. Rennes: Presses universitaires de Rennes, 2015.

Wang, Juan. “The Weight of Frivolous Matters: Shanghai Tabloid Culture, 1897-1911,” Ph.D. dissertation, History, Stanford University, 2004.

Yeh, Catherine Vance. “Shanghai Leisure, Print Entertainment and the Tabloids Xiaobao 小報.” In Joining the Global Public: Word, Image and City in Early Chinese Newspapers, 1870-1910, edited by Rudolf Wagner, 201–34. State University of New York Press, 2005.

Zhang, Zhen. An Amorous History of the Silver Screen: Shanghai Cinema, 1896-1937. Chicago: Univ. of Chicago Press, 2005.

Xie Guoxing 謝國興, ed. Wanxiang xiaobao jindai Zhongguo chengshi de wenhua shehui yu zhengzhi 萬象小報:近代中國城市的文化、社會與政治 (All-Seeing Tabloid Newspapers: Modern Chinese Urban Culture, Society and Politics). 台北 Taipei: 中央研究院近代史研究所 Institute of Modern History, 2013.