Introduction

Parental expectations are often a source of psychological distress for adolescents (Agliata and Renk 2009; Costigan et al. 2010). This is especially true for people from Asian cultures because family ties are particularly important (Fatima et al. 2016; Kim and Park 2006). Chinese parents place great importance on their children’s education and academic achievement (Chen et al. 2012; Costigan et al. 2010). According to Li and Wang (2004), Confucian ideas emphasize intellectual development, love of learning, and skill acquisition.

This cultural characteristic affects the style of Chinese parenting. Influenced by Confucian traditions, one important goal of parenting is the development of children’s characteristics of self-perfection (Li and Wang 2004). Children are expected to strive for achievements for their family (Huntsinger et al. 2000). Owing to the belief in family accountability, parents are believed to be ultimately responsible for teaching their child well (Chen and Luster 2002). When children do not put in hard work, or if they under-achieve, Chinese parents view them as signs of failure and take it upon themselves to force their children to try harder (Guo 2013).

The emphasis of Chinese parents on children’s proper behavior and academic achievement during their upbringing is best illustrated by the “tiger mother” debate, which was suggested by Chua (2011) as her parenting style as a Chinese mother. The strictly academic-orienting parenting style involves overriding children’s preferences to achieve academic excellence, which is in sharp contrast with most Western parents who prefer to let children develop freely. Guo (2013) saw this tendency as the belief among Chinese that future orientations and social images are associated with scholarly achievement. Besides as personal achievements, the distinct style of parenting might also be related to the patterns of social interaction among Chinese people. A low achieving child might make the parents “lose face”, which is defined as a respectable social image or respect from other people (Chang and Holt 1994). Face loss is a major consideration in Chinese relationships, and was suggested to be a much stronger predictor of relationship deterioration than in other cultures (Cardon and Scott 2003; Kam and Bond 2008). Viewed this way, children’s academic achievements are related to the aim of developing a desirable social image and preserving respect from others (Guo 2013). Accordingly, in a study of Hong Kong parents, competence and achievement were indeed the most frequently mentioned characteristics expected of the child (Ho and Kang 1984).

Therefore, it is not at all surprising that the push for achievement is a major source of pressure for Chinese children (e.g., Eisenberg et al. 2009). Given the expectations of Chinese parents, “doing poorly in school” is one of the major causes of parental punishment (Solomon 1971). In addition, Chinese parents tend to choose harsh physical punishments as techniques of control and discipline (Ho 1986). Researchers have found that authoritative styles of parenting are closely associated with Chinese cultural values, and that Chinese parents indeed demonstrate higher levels of physical coercion, shaming, and love withdrawal (Eisenberg et al. 2009).

On the other hand, Yang (1981) argued that Chinese people attach a great weight to the anticipated reactions of others—a trait traceable to the Confucian emphasis on interrelatedness (Bond and Hwang 1986). As individuals try to maximize the positive esteem they get from others (parents in this case) and avoid their disapproval (Fung 1999), adolescents might also act according to what they think their parents want. Thus, even if parents do not have any expectations toward their children, children would act in a similar way as though there were.

From the above discussion, it is suggested that academic achievement and fulfilling parental expectations are important determinants of family harmony among Chinese families. When a child underperforms, it is expected that family harmony and, as a result, adolescents’ perceptions of their family functioning, would suffer as parents show their dissatisfaction toward their children’s academic performance or displeasure at the potential of “losing face”. Adolescents should also view their family more negatively if they receive punishments (physical or not) for their poor academic performance. Various theoretical perspectives have been used to analyze the role of parents toward children’s development and education, including family capital (Bornstein and Bradley 2003; Ferreira et al. 2016), family investment (Conger and Donnellan 2007), and parental involvement (Epstein 1987; Fan and Chen 2001). However, little work has been done on this while also considering the gender dimension.

The relative social status of women in Chinese culture is well documented (Bond 2010; Schwartz and Rubel-Lifschitz 2009). While Chinese, like Americans, demonstrate typical gender stereotypes (e.g., which gender is more adventurous), sex-role differentiation is arguably more pronounced among the Chinese (Bond and Hwang 1986; Marshall 2008), which can be interpreted in terms of the traditional Chinese and male-centered family structure (Bond and Hwang 1986). In practice, Chinese parents do have different expectations toward their daughters and sons, and could even go as far as offering their sons preferential treatment in terms of education resources (Yu and Su 2006).

Boys’ and girls’ academic motivation also demonstrate very different patterns. In a study of grades 5 and 6 children in Hong Kong, Li (1970) reported that test anxiety in boys was related to high parental demands for strictness, obedience, and fostering dependency, while test anxiety in girls was related to low parental harshness and dominance. These results suggested that Chinese parents place much higher hope on the performance of boys over girls. The method of punishment or discipline might also differ across gender. Ho (1986) reported that parents tend to use more induction when disciplining daughters and rely more on love withdrawal and power assertion when disciplining sons. This was confirmed by Tang (2006), who found boys to experience higher rates and more frequent parent corporal punishment. The divergent parenting methods across gender is another potential mechanism from academic achievement to adolescents’ views on family. (Note that this does not necessarily mean that boys’ and girls’ perceived parenting style are different; see, for example, Cheung and McBride-Chang 2008).

Two hypotheses can be formulated about gender, parental expectations, and adolescents’ perceptions of family functioning. We label them “avoiding disappointment” and “fulfilling expectation” (from boys’ point of view). First, when boys perform well academically, their parents might not feel particularly overjoyed because their sons simply achieved what they were expected to, which would not affect the parents’ attitudes and thus the boys’ perceptions of family. In this case, the outcome merely prevents the parents from disappointment. On the other hand, when girls perform well, it exceeds the parents’ expectations and would positively relate to family harmony. Girls would then develop more positive perceptions of family functioning. The second possibility is that, when parents have high expectations for their sons, they would feel particularly satisfied when their expectations are fulfilled. They would in turn show more care and affection toward the boy, improving family harmony. When girls exceed their expectations, they would not change their attitude as they do not perceive academic performance as important for their daughters. Because we are not aware of any theoretical contributions that would lead us to favor either hypothesis, we believe this is largely an empirical issue.

Method

Participants

A total of 1830 students aged 10–19 from nine primary and secondary schools in Hong Kong participated in the survey. In the analysis below, a total of 1597 valid observations were used due to missing responses. Participants with missing responses in any of the questions were dropped, resulting in 233 unusable observations. An investigation of these observations showed that they were distributed roughly proportional to gender, academic achievement, and perceptions on family. Therefore we found it unlikely that this would introduce bias into our analysis. The final sample consisted of 52.1 % female and 47.9 % male, which was representative of the entire Hong Kong population (53.3 % female; see Census and Statistics Department 2012). The average age was 14.2 with a SD of 1.72. Although the age range of 10–19 was quite wide, there were not a lot of respondents on either extremes and the results were robust to the use of different age range, to be discussed below.

Procedure

We obtained ethical approval from a human research ethics committee from a Hong Kong university. The research was conducted through the assistance of a non-governmental social work organization and nine participating schools. The consent of the principals of the participating schools was obtained prior to the research. With the assistance of school personnel, information sheets outlining the research aims, usage of data, and voluntary participation of their children were circulated to all participating parents at least 1 week before the commencement of the survey. The surveys were conducted in the respondents’ own classrooms during school hours. The distribution and collection of surveys was done by social workers, and they were available to offer assistance such as clarifying some questionnaire items to the participants whenever necessary. All participants were informed of the purpose of the survey and the anonymity of their responses. Their participation was entirely voluntary (they could withdraw at any time), and the survey achieved an average response rate of 88 %.

Measures

Dependent Variable: Chinese Family Assessment Instrument (C-FAI)

This study used the C-FAI, developed by Shek and colleagues specifically for the Chinese cultural background (e.g., Shek 2002; Shek and Ma 2010). It consists of five dimensions (mutuality, communication, conflict and harmony, parental concern, and parental control), which are subsumed under two factors: family interaction and parenting (Shek and Ma 2010). This fits with the definition of family functioning either as dynamic interaction within family units (David 1978) or as the way in which a family fulfills its functions (Patterson 2002). The C-FAI has high reliability, validity, and is significantly correlated with other measures of family functioning and individual psychological well-being (Shek and Ma 2010). Respondents were asked to evaluate, on a 5-point scale, how much they think the statements resemble their own family (from Very similar to Very different). Examples include “Family members support each other” and “Parents often talk to children.” The responses were then summed up to produce an index ranging from 33 to 165. Higher scores represented higher level of dysfunction in family functioning. In the current study, the reliability for this scale was high with a Cronbach’s alpha of 0.96.

Independent Variables: Gender and Academic Achievement

Gender was coded as a dichotomous variable (1 = Female) in this study. Two questions were asked in the survey to measure academic achievement: respondents’ evaluation of their own academic performance and their satisfaction with it (both on a 5-point Likert scale). We created an index of academic achievement by summing both responses. Scores ranged from 2 to 10, with higher scores representing greater academic achievement. Since both questions captured the respondents’ own evaluation, we found it appropriate to create a composite measure, which gives a more complete picture of their academic performance and also minimizes error associated with each response. Nevertheless, the correlations between the combined measure and each item were higher than 0.9.

Control Variables

Although all respondents could be considered as adolescents, their age range was still considerable (10 to 19). Scholars have suggested that Chinese parents tend to impose harsher discipline upon older children (e.g., Ho 1986). To capture the effect of age, respondents’ age was used as a control variable. To further assess the potential impact of age range, several robustness tests were performed. The main results were found to hold even if we drop the older respondents (above the age of 16, about 100 observations), the younger ones (below the age of 14, 400 observations), or simply working with a narrower age range from 14 to 16 (while still retaining two-thirds of the observations). Further tests showed that the results were also not sensitive to the choice of the cut-off age (results available from the authors upon request).

Since respondents’ family characteristics and family structure should be the most important determinant of their views on family, they had to be accounted for. In the analysis below, a dummy variable was constructed to represent the marital status of the respondents’ parents (1  = married; 0 = separated/divorced/other). Controls were also included for whether respondents live with their parents and whether they have siblings.

To ensure robustness of our results, in addition to the base model above, two alternative models were estimated. The first model controlled for additional personal demographics. Ethnicity was an obvious candidate to control for, given the arguments relating to the Chinese family. Alternatively, omitting non-Chinese respondents did not change the results. Also, respondents’ religious background might also change their views on family. Dummy variables were used to capture ethnicity (1 = Chinese; 0 = Southeast Asian/Indian/Middle Eastern/Other) and religious background (1 = Have a religion; 0 = No religion). Creating different dummy variables for each religious belief (Catholic/Christian/Buddhist/Islamic/Other) did not affect the results.

To take into account the fact that variation in family satisfaction might come from life satisfaction in general or one’s positive feeling about oneself, the results below were also tested against the inclusion of life satisfaction and self-esteem as controls. The Satisfaction with Life scale developed by Diener et al. (1985) and the Rosenberg Self-Esteem Scale (Rosenberg 1965) were used. Both of these scales were suggested to be applicable to, and widely used with, Chinese (e.g., Chui and Chan 2012; Chui and Wong 2016). Higher values represented higher life satisfaction and self-esteem. These two scales had high reliability in our survey, with Cronbach’s alphas of 0.82 (life satisfaction) and 0.88 (self-esteem). Brief summary statistics and correlations between the major variables can be found in Table 1.

Table 1 Descriptive statistics, Cronbach’s alpha, and correlations
Full size table

Data Analyses

The objective of the statistical analyses below was to test the relationship between gender, academic achievement, and adolescents’ perceptions of family functioning. Gender and academic achievement were entered into a simple OLS model first (base model). This specification estimated the average effect of both variables. However, according to our theoretical arguments, the conditional effect can only be captured when the interaction term (gender x academic achievement) is inserted (interaction model). Next, two further sets of controls discussed above were used to ensure the robustness of our results. The demographic model included all variables in the interaction model plus ethnicity and religion as additional controls. The self model also built on the interaction model, but included self-esteem and life satisfaction instead. Finally, the subscales of the C-FAI would be used to test the mechanisms of the effects.

Results

Empirical results largely supported the avoiding disappointment hypothesis. Table 2 shows the OLS regression results. The base model included gender, academic achievement, and key control variables to explain adolescents’ perceptions of family functioning. Because higher C-FAI scores represented a higher level of dysfunction in family functioning, it can be seen that girls held a more positive view toward family, all else equal. The same was true for students who had good academic achievement, as well as those with married parents. These results were all rather straightforward and consistent with expectations.

Table 2 OLS regression models of adolescents’ views on family
Full size table

However, the base model did not allow for the differential effect of academic achievement across gender. Divergent patterns could be identified when we allow for this dynamic by including an interaction term between gender and academic achievement, as shown in the second column (interaction model). Following Brambor et al. (2006), after interaction terms are introduced, the significance levels of individual variables are no longer the overriding concern. Instead, F-tests for academic achievement and its interaction with gender showed that they were jointly significant (Kam and Franzese 2007), confirming the interaction effects between gender and academic achievements (F = 61.17, p < 0.001). The last two columns of Table 2 tested this result against the inclusion of two other groups of variables. Although the magnitude and significance of the coefficients dropped in the model with other mental health indicators (self-esteem and life satisfaction), this is understandable given the very strong effect of the two indicators on family views. Still, the interaction term between gender and academic achievement remained significant.

Models with interaction terms cannot be easily interpreted without a graphical presentation (Brambor et al. 2006), which can be seen in Figure 1. The graph estimated for the interaction model was very similar to the demographic model and thus was not replicated here to save space (available from the authors upon request). According to the estimations of the two models, the lines represented the predicted C-FAI scores of boys and girls when their academic achievement was high or low, respectively. This prediction followed the methodology of Lai (2009). Low/High (of academic achievement) represented values one standard deviation below/above the mean. All other variables were held at their respective mean. The overall picture was consistent across the two models. While perceptions of family functioning were at a similarly low level for both genders when the respondent was not doing well academically (although girls still held a slightly more positive perceptions), the change associated with an increase in academic achievement was very different.

Fig. 1
figure 1

Predicted C-FAI scores by gender and academic achievement (higher scores = worse functioning). Note: Simulation via CLARIFY (Tomz et al. 2003) based on the model specifications in Table 2 while holding other control variables at their means

Full size image

For example, in the demographic model, a change in academic achievement from Low to High was associated with a 14-point improvement in the C-FAI score for girls (from approximately 85 to 71, a 16.5 % decrease). However, the change was just about 8.4 points for boys (a 9.8 % decrease). Although the contrast based on the predicted values from the self model was less pronounced, the picture was no less remarkable: Moving academic achievement from Low to High came with an improved girls’ views on family, but the value was virtually constant for boys (a predicted change from 83.6 to 82.6).

Parental Expectations across Gender: Evidence from the C-FAI Subscales

What actually drove the divergent pattern for boys and girls in our results? To further buttress our statistical analysis and theoretical arguments, we looked for evidence by using the subscales of the C-FAI. As discussed above, Shek and Ma (2010) reported five dimensions in the C-FAI, which can be categorized under two factors: family interaction (mutuality, communication, and conflict and harmony) and parenting (parental concern and parental control). With these measures, it was possible to test the different dimensions of family functioning and investigate if the patterns conformed to our story.

Dividing our respondents into four groups (by gender and low/high academic achievement), we compared the average value of the five dimensions of C-FAI. Academic achievement cut-off was at mean. The five subscales were standardized (with mean = 0 and SD = 1) to make them comparable because they consisted of different numbers of questions. Figure 2 showed two very different patterns depending on the subscale. The distinctive nature of the conflict subscale stood out. Not only did it demonstrate a unique pattern where the change in family functioning associated with academic performance for both gender was very similar, but the score for boys actually surpassed girls for the only time in this research. With the caution of causality in mind, this could mean that when adolescents of both genders did well academically, there was also a greater family harmony by avoiding conflict—possibly the parental punishment that would occur in the event of underachievement.

Fig. 2
figure 2

Average C-FAI subscale scores by gender and academic achievement (higher scores = worse functioning). Note: Simulation via CLARIFY (Tomz et al. 2003) based on the interaction model while holding other control variables at their means

Full size image

Mutuality, communication, and parental concern demonstrated a pattern quite similar to Fig. 1, with the exception of the parental control subscale. For boys, better academic achievement was not associated with lower level of parental control (the figure was even slightly higher), while the pattern for girls was similar as the other dimensions. This could be interpreted as the fact that the extent of parents’ control on boys was not related to how well they did at school. This provided further evidence for the recurring idea in this study that Chinese parents are stricter toward their sons (Ho 1986; Tang 2006).

Summarizing the results, they provided prima facie evidence for our argument that parental expectations about academic achievement across gender were the underlying causal mechanism for the divergent effects. While boys (like girls) should not be subject to punishment if they do well at school, parents show less compassion and care toward them as compared to girls who also do well at school. We argued that the subscale scores of the C-FAI fit with our causal story of parental expectations and the “avoiding disappointment” hypothesis. Of course, such a conclusion is preliminary and suffers from some limitations, to be discussed below.

Discussion

The relationship between gender, academic achievement, and family functioning is not currently well understood. Focusing on Chinese families, we derived two competing hypotheses and tested them empirically. Because parents are more likely to have higher expectations toward their sons, they could either be overjoyed if their expectations are fulfilled (“fulfilling expectation”) or simply feel that their disappointment is avoided (“avoiding disappointment”). These would then translate into family harmony and adolescents’ views on family. Our results lent support to the latter avoiding disappointment hypothesis: Sons who did well in Chinese families only prevented parents from disappointment; parents were actually happier when they had high achieving daughters because of the lower initial expectations. This result adds a previously overlooked gender dimension to the “tiger mother” image of Chinese parents.

Overall, girls held a more positive outlook toward their family. However, after an interaction term was included, gender was no longer statistically significant. This did not mean that gender was not a crucial factor. The demographic model showed that, while girls on average had a more positive view of their family than boys, if we took academic achievement into account (that is, comparing high-achieving boys with high-achieving girls), high-achieving boys had more positive views than low-achieving girls. Again, because the coefficients only represented an average effect, it is not surprising that the coefficient for gender was statistically insignificant in some of the models.

While it is traditionally believed that parents have higher expectations for boys, in a recent study of Macau college students, Found and Sam (2013) did not find any gender difference in parental expectations. As our research did not assess parental expectations directly, there was no way to compare our results with theirs. However, even if Chinese parents do not have different expectations for sons and daughters, our results reflected the fact that they might have different reactions toward their achievements. Our conclusion might be challenged if Chinese parents instead hold higher expectations toward daughters, which was quite improbable given our understanding of Chinese families and a large literature suggesting otherwise.

For researchers in the field, our results should serve as a call to more careful investigation of the subtle dynamics behind analytical models (which focus on average effects), as divergent patterns like the one we uncovered could easily be overlooked. Our findings also carry implications for practitioners. To the extent that adolescents’ assessment of family functioning actually reflects their parents’ reactions (an assumption we will discuss next), boys benefited less from a higher academic achievement than girls do. Parents should be encouraged to adopt equal expectations toward their sons and daughters in order to avoid having different reactions to their achievements, which could adversely affect how boys view the family.

We conclude by discussing the limitations of the study. First, we did not assess family functioning and parental reactions from the parents’ point of view. Family functioning was captured based only on the assessment of the adolescent. This could increase the level of uncertainty about the causal mechanism. In this case, our results might be reflecting the pressure boys had on themselves (having good results was not considered to be enough to making parents happier), even if parents did not actually have biased expectations or react differently based on gender. It could also be possible that children who do well academically tend to perceive their mothers as more demanding and restrictive (Cheung and McBride-Chang 2008). Although the close relationship between academic achievement and subjective well-being is well-established in adolescent studies, a more convincing result can be achieved if we can ideally capture the views of both the adolescents and their parents.

Another weakness was the use of self-reported data for academic achievement. The choice was primarily practical, as there is no centralized public examination in Hong Kong until around the age of 17 or 18. A comparable assessment of academic results for all respondents was difficult to be obtained. Furthermore, as respondents were recruited from nine different schools, comparable measures of their academic results (such as school exams) were simply not available even among those of similar age. Another justification came from Xiao et al. (2009), who found that academic satisfaction was directly influenced by academic results. In any event, the findings of this study would be strengthened if they can be replicated by objective and comparable indicators of academic performances.

Finally, as this study was a cross-sectional one, it might be the case that the causal direction ran in the opposite direction. For example, Stanard et al. (2010) argued that family closeness could lead to better academic results because it provided an optimal environment for adolescents to excel in. While this notion is intuitively sound, it could not sufficiently explain the empirical pattern uncovered in this study (it could imply that boys required less family harmony to do well academically). Still, we hasten to add that our theoretical arguments and empirical results did not preclude the possibility that the link between family and academic achievement ran in both directions. A number of studies indeed highlighted the impact of environmental factors (including family) on children’s academic achievements (e.g., Cheung and McBride-Chang 2008; Fan and Chen 2001; Stanard et al. 2010). It is acknowledged that this question cannot be resolved completely by the current study and must be left for future research.