Who joins which network, and why?

Abstract

Social networks—such as alumni networks—affect the productivity of highly educated immigrants through collaboration, knowledge spillover, job referrals, etc. Using online social networking data, we study how international graduate students choose an online identity associated with the “main alumni network” when there are multiple schools attended, but only one alumni network can be selected. We focus on students who receive undergraduate education in China and graduate education in the USA using data from Renren.com, a China-based Facebook-type networking website, on which users can only identify one network among multiple options. We find that pre-migration preferences for Western culture—measured by English-name usage—increase the likelihood of choosing the US school network. Causal inference is based on the design of linguistic instrumental variables for English-name usage. While school prestige should also affect the choice of alumni network, our results highlight the importance of cultural preferences in network choices: The effect of English-name usage is significant regardless of school rankings.

Appendices

Appendix A: Name Heterophone

In the empirical analysis, we use the indicator of the heterophonic Chinese name to instrument for English-name usage. This is based on the idea that a student whose Chinese name is pronounced differently in Chinese and English should be more likely to use an English name in real life and on the social network.

Chinese is a logo-syllabic language and has a unique writing system. Since the nineteenth century, many linguistic systems have been invented for Romanizing Chinese characters using the Roman alphabet. The current official Romanization system is the pinyin system, which was invented in the mid-1950s. Unlike earlier Romanization systems (e.g., the Wade–Giles system), the pinyin system was designed mainly based on linguistic attributes of Chinese, which helped reduce the illiteracy rate in China at that time. However, due to the long linguistic distance between Chinese and English (Crowley and Bowern 2010), this Chinese-based pinyin system cannot always perfectly reflect pronunciation rules of Chinese in English (Bassetti 2007).

Table 9 Heterophonic Linguistic elements in Chinese

In Table 9, we present a list of heterophonic linguistic elements in Chinese. In Column 1, we show these elements. In Column 2, we present examples of characters that contain these linguistic elements. In Column 3, we show the English words that approximate to how the character examples in Column 2 are pronounced in English. In Column 4, we show how these characters should be ideally pronounced.

We first list consonant-based heterophonic elements. The pinyin system generally uses one letter to represent a consonant element, which causes heterophone due to different pronunciation rules in Chinese and English (Lee and Zee 2003). On the contrary, some consonant elements are Romanized into two letters in the Wade–Giles system. For example, c- (followed by, e.g., -an) is pronounced as k- in English and ts- in Chinese. Note that c- is Romanized as ts- in the Wade–Giles system, hence in this case, c- is heterophonic in the pinyin system but is probably homophonic (as ts-) in the Wade–Giles system.

In the rest of the table, we list vowel-based heterophonic elements. In Chinese characters, these vowel-based elements are usually led by consonants. One major type of heterophone involves Chinese vowels that are heterophonic even within Chinese (Lee and Zee 2003). For example, -i is a heterophonic vowel in Chinese: Its pronunciation rule appears to be different in, say, ti and si, but -i is homophonic in English (which is pronounced similar to -i in ti in Chinese). Another type of heterophone involves the difference between the velar nasal and the alveolar nasal (e.g., Zee (1985)) in Chinese. These heterophonic elements include -ang and -eng.

Appendix B: Name Heterophone in External Samples

The classification of heterophonic name has two concerns: first, whether the classification rule is robust and second, whether students who move abroad are less likely to have heterophonic Chinese names. In Table 10, we use external samples to study these concerns.

Table 10 Heterophonic names in external data

We use data from two schools that publish students’ post-graduation outcomes, namely China Center for Economic Research (at Peking University) and Nanjing Foreign Language School. These samples are not online social networking samples. Table 10 shows that the percentage of heterophonic Chinese names among students who stay in China is 42.4%, while this percentage is 43.3% among students who leave China after graduation. The difference is small and insignificant. This suggests that name heterophone should not be a special attribute among students who move abroad. In addition, the percentage of name heterophone in Table 10 (approximately 43%) is very close to that in the social networking sample used in this paper. This indicates that the classification rule of name heterophone is also robust to changes to sample.

Appendix C: Logit regression results in Sect. 4.1

In Sect. 4.1, we report the results of OLS regressions. In Table 4, we present evidence that English-name usage is positively correlated with the identification of the US school alumni network. Note that the dependent variable, i.e., the indicator of identifying the US school alumni network, is a binary variable. To re-examine the empirical specification using the nonlinear model, in Table 11 we run logit regressions of the choice of US school alumni network on English-name usage and report the marginal effects. Results show that the marginal effects of English-name usage estimated using logit are numerically very close to OLS estimates reported in Table 4.

Table 11 The effect of English-name usage on the identification of the US school alumni network: logit regressions