Has socioeconomic development reduced the English divide? A statistical analysis of access to English skills in Japan

ABSTRACT

As English skills are being increasingly perceived as valuable human capital, an awareness of inequitable access to the acquisition of this ‘profitable’ language has developed. Although this allegedly unequal access, which can be referred to as the ‘English divide’, is a matter of concern to applied linguists, little research has been conducted on the historical transformation (or maintenance) of the English divide within a specific society. Therefore, this paper examines a historical change in the English divide in Japan as a highly industrialised country through statistical analysis of a large, population-based sample. Results revealed that, with Japan attaining various socioeconomic developments such as reduced income gaps and improvements in educational conditions, some forms of the English divide originating from material gaps have gradually decreased. In contrast, other forms of the English divide that are led by cultural disparities have not decreased at all. Based on these findings, I discuss the inevitable difficulties of tackling the English divide and suggest some possible solutions for creating substantive, rather than formal, equality in English language education.

KEYWORDS:

• English divide
• unequal access to English language skills
• pre- and post-war Japan
• statistical analysis
• social survey
• language policy and planning

Acknowledgements

I thank the Japanese General Social Surveys (JGSS), which provided me with valuable data sets. JGSS are designed and carried out by the JGSS Research Center at Osaka University of Commerce (Joint Usage / Research Center for Japanese General Social Surveys accredited by Minister of Education, Culture, Sports, Science and Technology), in collaboration with the Institute of Social Science at the University of Tokyo.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 The self-reported nature of these English proficiency variables does not seem to be problematic for the purpose of this paper based on the following rationales in Terasawa (2015, 32–33). First, this paper does not focus on an absolute number of respondents who answered to have English proficiency, but a relative or comparative aspect by estimating gaps of their responses between particular groups. Second, self-assessed English skills based on concrete behaviours such as can-do statements (the English proficiency variables in this paper are ones of them) are found to be well correlated with English proficiency measured by objective language tests.

2 These job categories are based on an eight-category framework conventionally used in Japanese labour study. These categories are as follows: (1) professional; (2) managerial; (3) clerical; (4) sales; (5) skilled; (6) semi-skilled; (7) unskilled; and (8) agriculture, forestry, and fishery. This paper considered the first two categories as upper white-collar occupations.

3 Although this paper regarded and analysed this variable as an interval variable for convenience, strictly, it is an ordinal variable. However, considering it an interval variable probably did not cause major problems because a similar result was obtained by a previous work that analysed this variable as an ordinal one (Terasawa 2015, Chapter 2).

4 It should be noted that these percentages were calculated based on the total number of secondary school students rather than a same-age population. Considering the pre-war situation, in which many elementary school graduates could not proceed to secondary schools, the actual percentage of English learners in this generation must be lower than 30%.

5 In addition to these variables, I have estimated percentages of English speakers by considering mother’s education level, father’s job, and rurality/non-rurality of the individual’s hometown, as well as the percentages of Japanese who answered that they had English reading abilities (Terasawa 2015, Chapter 2); thus, I have confirmed that they did not affect the conclusion drawn.

6 In the figure, a rectangle represents an observed variable while an ellipse represents a latent variable; a circled e represents a measurement error and a triangle with a number ‘1’ inside represents a constant. Six straight paths to English proficiency presuppose a causal relationship between two variables, six straight paths from a latent variable to an observed variable are factor loadings and a curved path represents a correlation.

7 Note that I included several analytical presuppositions in this model: (a) an equation constraint to every factor loading (i.e. each factor loading is assumed to have an identical value, irrelevant of cohort groups); (b) fixed values to 1.00 for the variances of the three latent variables; and (c) fixed values to 0.00 for the constants of the three latent variables. These presuppositions can be justified by the fact that the historical changes in these values are of little interest to this study.

8 Considering various fit measures obtained from the analysis (χ²(230) = 3736.831 (p < .001), RMSEA = 0.120, GFI = 0.911, CFI = 0.897, SRMR = 0.321), deeming this model as well fitted to the data is a challenge. This is probably largely attributed to the fact that I excluded from the model some variables that possibly have strong explanatory power for variations in English proficiency, such as respondents’ education levels, their experiences in using English, and current job status. Indeed, including these variables in the model dramatically improved these fit measures, but I decided to maintain the original model because the paper’s primary purpose was to reveal how socioeconomic factors influenced English proficiency rather than propose a model that successfully explained variances in English proficiency.

9 To improve readability, measurement errors (represented by a circled e in Figure 2) and constants (a triangle) were omitted from Figure 3.

10 Although not directly related to this paper, the rapid decrease in the male advantage in obtaining English proficiency is noteworthy. Analysis showed that ‘Male’ in C1 significantly affected English proficiency, and, more importantly, the size of this effect (non-standardised estimate = 0.75; standardised estimate = 0.31) was much greater than most other factors. In C2, however, the effects of ‘Male’ considerably diminished and almost disappeared in C3 or the younger cohorts, thus suggesting that the substantive gender inequality in English language learning that once existed in pre-war Japan has largely improved over generations. This implies that a gender gap in the English divide can be completely different in quality from other forms of the English divide.

11 Percentages represented are the sum of the percentages for ‘I agree’ and ‘I somewhat agree’.

Additional information

Funding

This work was supported by Japan Society for the Promotion of Science KAKENHI [grant number 14J10080].