In Western societies, alcohol use among university students has been extensively studied for many years, and gender is one of the most frequently discussed correlates. [7] summarized the findings of early American studies, drawing the conclusion that, in general, men drink more frequently, consume larger quantities of alcohol, are more likely to drink to get drunk, and experience more negative consequences from drinking. Recent studies in other Western countries have found similar gender differences in drinking frequency ([9]; [10]; [17]), drinking quantity ([17]), and problem drinking ([9]; [17]). However, [7] suggested that this common finding may result from a somewhat false dichotomy. The difference in alcohol consumption between most men and women is not significant. Instead, the extreme, heavy drinking patterns of a small proportion of men are the real cause of gender differences in the average ([7]). Compared with female university students, male university students have significantly higher rates of heavy episodic drinking ([9]). Gender differences in abusive drinking, in fact, are greater than gender differences in drinking frequency ([17]). In Hong Kong society, findings related to gender share many similarities with those in Western societies. Binge drinking is significantly associated with male gender ([11]; [15]; [16]). More men than women reported ever drinking ([2]; [11]). [2] believe that gender differences in substance use may be related to beliefs and sociocultural factors, which discourage women’s substance use.

Age is another common correlate, but the findings of previous studies are inconsistent. For example, in a study conducted in Belgium, [17] found that older students were less likely to drink and engage in abusive drinking. This may stem from a protective maturing effect of age. As [7] summarized, after advancing to higher grades, men may learn a more controlled and less destructive way of drinking. However, a study conducted in Brazil came to the opposite conclusion, finding that older students had a higher likelihood of drinking and binge drinking ([10]). The relationship between age and drinking behaviors is difficult to determine clearly, as the protective maturing effect is likely to be covered by other factors. [17] argue that the time spent in university is one such risk factor. Although some studies have both found that older students are more likely to binge drink ([14]; [16]), it is unclear whether this conclusion represents a definitive and universal relationship between age and drinking patterns.

The two main categories of students at university in Hong Kong are local students and students from the Chinese Mainland, but few studies have focused on differences in drinking behaviors between the two groups. [11] categorized students according to their place of birth and found that male students from the Chinese Mainland had a relatively high rate of alcohol consumption. They explained this difference by arguing that Hong Kong students live with their families, which means they have less chance of accessing alcohol ([11]). However, [11] did not conduct a comprehensive analysis of differences in drinking behaviors, and there may be other ways to explain the possible differences. In addition, there are many non-Chinese students in Hong Kong’s universities, but very few studies focus on them.

When discussing university students’ drinking behaviors, religion is a correlate that is rarely mentioned. A study on Hong Kong adolescents’ religious beliefs shows that 34.3% of students have religious beliefs. Among Non-Chinese Speaking South and Southeast Asian Students (NCS), the proportion with religious beliefs is significantly higher than that of Chinese Mainland and Hong Kong students ([28]). The major religious beliefs vary among different student groups. For NCS, the major religions are Islam and Christianity, while for Chinese Mainland students and Hong Kong students, the major religions, respectively, are Buddhism and Christianity ([28]).

Compared to students with religious beliefs, students without religious beliefs have significantly higher levels of drinking frequency, quantity, and getting drunk ([21]). Different religions also have different attitudes towards alcohol use. According to [19], Buddhists have the most positive attitudes towards alcohol use. Next are Christians. Muslims are the last. Many international students in Hong Kong come from countries where Islam is the dominant religion. In Islam, alcohol use is strictly prohibited ([3]). Therefore, the overall prevalence of binge drinking among university students is much lower in Muslim countries than in Buddhist countries. These differences can be partially explained by students’ religious beliefs ([26]). In addition, higher levels of religious commitment have a protective effect on alcohol use ([3]). Among students with a higher level of involvement in religious activities, the prevalence of binge drinking is much lower ([26]).

Current research has not reached a consensus on the relationship between parental socioeconomic status and university students’ drinking behaviors. The first debate in the existing research is about whether there is an association between parental socioeconomic status and university students’ drinking. Although many studies have confirmed the existence of an association ([9]; [10]; [12]; [14]; [18]), there are still studies with different findings, suggesting that mothers’ educational attainment and family’s financial situation are not associated with alcohol use ([2]), and that the relationships between parental educational attainment and alcohol consumption are not significant ([9]).

The second debate is about what kind of relationship exists between the two. Some studies have found that university students with higher SES have higher drinking rates ([14]), drink more frequently ([18]), and are more likely to face alcohol-related problems ([12]). Others argue that individuals with low SES are most likely to engage in binge drinking ([10]), and that students who feel they do not have sufficient income are more likely to drink and be classified as problem drinkers ([9]).

Accommodation type, which refers to students’ specific residential arrangements during the academic semester (e.g., residing with parents, living in university dormitories, or renting independently off-campus), is an essential sociodemographic factor frequently discussed in the analysis of university students’ alcohol use. These living arrangements have been consistently linked to various drinking indicators. For instance, [9] found in a UK study that living with parents is significantly and negatively correlated with high-frequency drinking, heavy episodic drinking, and problem drinking. Similarly, [10] demonstrated in a Brazilian study that university students who do not reside with family (e.g., living alone or with peers) are more prone to engaging in binge drinking.

Empirical literature supports the mechanism underlying these variations, suggesting that the transition away from the parental home significantly alters social control structures. Specifically, longitudinal research by [25] and [1] demonstrates that relocating away from family drastically reduces direct parental monitoring and expands personal autonomy. This reduction in parentally enforced behavioral boundaries, combined with increased proximity to independent living environments, leaves students substantially more vulnerable to peer drinking norms and local campus alcohol cultures.

Besides, variations in institutional management systems can lead to divergent outcomes. For instance, [14] discovered that due to the stringent dormitory regulations in Chinese mainland universities, off-campus students exhibited higher binge drinking rates than their on-campus counterparts. In Hong Kong, empirical evidence aligns more closely with the Western paradigm. [16] indicated that full-time residential hall living is independently associated with past-month binge drinking, while [2] observed that university students who had ever consumed alcohol were more likely to reside in university hostels compared to lifetime abstainers.

This is a cross-sectional study, which employed quantitative methods to identify key sociodemographic factors associated with Hong Kong university students’ drinking behaviors. Full-time university students from a major public university in Hong Kong, aged 18 to 24, were recruited using convenience sampling. Data were collected through an online survey, and the questionnaire was distributed via QuestionPro. A total of 204 questionnaires were started, and 155 completed questionnaires were returned (response rate: 75.98%). 147 participants met the criteria (eligibility rate: 94.84%). The average completion time for the questionnaire was 7 minutes. The final sample size was 147.

This study was approved by the Research Ethics Committee at this major public university in Hong Kong. Participants were provided self-administered questionnaires, and each questionnaire included an informed consent statement. Participation was voluntary, and students could withdraw at any time without penalty. All responses were collected anonymously. Data were securely stored on password-protected devices and only accessible to the research team.

For sociodemographic characteristics, information was collected about gender, study year, age, student type (local, mainland, international), accommodation type (on-campus, off-campus), religion (Buddhism, Christianity, Islam, none, other), mother’s and father’s education (primary or below, junior high school, senior high school, university or above), and perceived family financial situation (very poor, tight, comfortable, well-off). The Alcohol Use Disorders Identification Test (AUDIT) was adopted in this study to measure participants’ alcohol use. This is because the AUDIT not only demonstrates high sensitivity and specificity among self-report screening measures, but has also been proven to be highly effective in identifying hazardous alcohol use or alcohol use disorders among university students ([22]). The AUDIT consists of ten items, which cover three domains. They are hazardous alcohol use, dependence symptoms, and harmful alcohol use ([5]). It assesses people’s alcohol use over the past year. The score range for the AUDIT is 0 - 40 in this study.

The third item in the AUDIT is about the frequency of heavy drinking. We made some adjustments to this question. The original question was “How often do you have six or more drinks on one occasion?” We changed it to “How often do you have: For females: 4 or more drinks on one occasion; For males: 5 or more drinks on one occasion?” This adjustment was based on the gender-specific definition of binge drinking. According to the [20], binge drinking is defined as a pattern of consuming five or more drinks for males, or four or more drinks for females, in about two hours. Except for the third question, all other questions in the AUDIT remain unchanged. However, because the third item was adjusted in this study, the standard AUDIT risk categories (e.g., low-risk range, hazardous or harmful range, alcohol dependent range) may not apply ([4]). For this reason, the AUDIT was treated as a modified screening instrument in this study, and only the continuous AUDIT score was used to reflect the severity of alcohol-related problems.

In addition, this section also collects three dimensions of frequency information from participants who report drinking. The first dimension is about the frequency of drinking for different reasons. The second dimension is about the frequency of drinking with different groups of people. The third dimension is about the frequency of drinking on different occasions. For the first dimension, we used Cooper’s (1994, as cited in [23]) classification of drinking motives. The four categories are enhancement (Because you like the feeling), social (Because it helps you to enjoy a party), coping (To forget your worries), and conformity (Because your friends pressure you to drink). We revised the wording in the first category. We replaced “enhancement” with “enjoyment” to make it easier for participants to understand.

All statistical analyses were performed using the package SPSS Version 29. The significance level was set at p < 0.05. Continuous variables were described as means and standard deviations, and categorical variables were described as absolute values and percentages. Frequency tables were generated, and stacked bar charts were made. The bivariate analyses were then conducted to examine the associations between sociodemographic variables and drinking behaviors, including drinking status, binge drinking behaviors, number of standard drinks, and AUDIT score. For binge drinking behaviors and AUDIT score, 27 participants who reported never drinking alcohol were classified as never having engaged in binge drinking, with a 0 AUDIT score. For number of standard drinks, only 120 participants who had consumed alcohol in the past year were included in the analysis. Chi-square tests, independent-samples t-tests, One-way ANOVA, and Pearson correlation were applied depending on the nature of the variables.

To further examine correlates of Hong Kong university students’ drinking behaviors, two multivariable logistic regression models and four multivariable linear regression models were constructed. However, both the number of standard drinks (skewness = 2.50) and AUDIT score (skewness = 1.38) showed positive skewness in the raw distribution. Therefore, before conducting bivariate and multivariate analyses, the number of standard drinks was log-transformed (log[x]), and AUDIT score was log-transformed (log[x + 1]). This transformation largely reduced skewness (number of standard drinks: skewness = 1.40; AUDIT score: skewness = −0.078).

Table 1includes descriptive statistics for study variables. Of the participants, 65.3% were female; 68.7% were junior students; 55.8% were local students. On average, students were aged 19.6 (std = 1.69). Most students had no religious beliefs (81.6%) and perceived their family’s financial situation as comfortable (76.9%). Slightly over half of the students had a mother and father whose education was at university or above (51.0% vs 55.1%). 49.0% of students lived on campus, while 51.0% lived off campus.

In addition, Table 1 also shows participants’ drinking behaviors. Of all the participants, 27 reported that they had never drunk alcohol in the past year, while 120 reported drinking alcohol with different frequencies. Most students reported their drinking frequency was monthly or less (40.1%). Besides, more participants reported never engaging in binge drinking behaviors than those who had (55.8% vs. 44.2%). Only 5.4% of participants engaged in binge drinking on a weekly or daily basis. Participants’ average AUDIT score was 3.97 (std = 4.02). Of 120 participants who reported alcohol use over the past year, on average, the number of standard drinks participants had on a typical day was 2.72 (std = 2.38).

Figures 1-3 present the drinking frequency of these 120 participants who have consumed alcohol over the past year by different motives, companions, and occasions. For drinking motives, social and enjoyment were the most common reasons for drinking. 39.2% and 43.4% of participants selected “often” or “almost always” for social and for enjoyment, respectively. On the contrary, conformity was the least common motive. Most people (85.9%) reported they rarely or never drink for this reason. For coping, 55% of participants reported a low frequency, selecting “never” or “rarely”, while the others chose a higher frequency. The number of participants reporting a low frequency was slightly higher than those reporting a medium-to-high frequency (Figure 1).

Table 1. Frequency distributions and descriptive statistics of variables.

Note: n = 147 except for the number of standard drinks, in which case n = 120.

Figure 1. Stacked bar chart for drinking motives (n = 120).

For Drinking companions, participants reported that they drank most frequently with friends and least frequently with strangers. The frequency distribution for drinking with fellow students was relatively similar to drinking alone. The number of participants reporting a low frequency was close to that reporting a medium-to-high frequency. Besides, most participants indicated that they did not drink with family members very frequently (Figure 2).

Figure 2. Stacked bar chart for drinking companions (n = 120).

Figure 3. Stacked bar chart for drinking occasions (n = 120).

For drinking occasions, participants reported that they drank most frequently in social gatherings with friends and least frequently in business or formal events. The frequency distribution for drinking in clubs or parties was relatively similar to that of drinking alone at home. The number of participants reporting a low frequency was slightly higher than a medium-to-high frequency. Few people reported frequent drinking during the dining period or family gatherings (Figure 3).

In bivariate analyses, as shown in Table 2, study year (χ² = 4.18, p = 0.041) and student type (χ² = 14.63, p < 0.001) were significantly associated with drinking status. There was no significant association between drinking status and gender, age, accommodation type, religion, mother’s and father’s education, and perceived family financial situation. The percentages of local students (82.9%) and mainland students (88.7%) who drank were significantly higher than those of international students (41.7%). The percentage of senior year students (91.3%) who drank was slightly higher than that of junior year students (77.2%).

Table 2. Descriptive statistics and bivariate associations with drinking status and binge drinking behavior.

Note: n = 147 except for gender, in which case n = 146 due to 1 person being third gender. Parental education, perceived family financial situation, and age are all not statistically significant with respect to either outcome. ^*p < 0.05.

An adjusted logistic regression was then conducted, controlling for six factors. As shown in Table 3, student type remained significant (p = 0.002). Compared to local students, international students had significantly lower odds of being past-year drinkers. Specifically, international students showed a dramatic reduction in drinking participation, with the risk decreasing by 97% (OR = 0.03). In addition, accommodation type (p = 0.021), which was not significant at the bivariate level, emerged as a significant correlate in the adjusted model. Living off-campus was significantly associated with a lower likelihood of drinking than living on-campus, with risk decreasing by 79% (OR = 0.21). After adjustment, study year lost its significance, and age, gender, and religion remained non-significant correlates (Table 3).

Table 3. Multivariable adjusted models for drinking status and binge drinking behavior.

Note: n = 146 due to 1 person being third gender. ^*p < 0.05.

For binge drinking behavior, as shown in Table 2, gender (χ² = 5.59, p = 0.018) and religion (χ² = 4.49, p = 0.034) were significantly associated with it in bivariate analyses. There was no significant association between binge drinking behavior and age, study year, student type, accommodation type, mother’s and father’s education, and perceived family financial situation. The percentage of males (58.0%) who had engaged in binge drinking behavior was significantly higher than that of females (37.5%). The percentage of non-religious students (48.3%) who had ever engaged in binge drinking behavior was significantly higher than that of religious students (25.9%).

After adjusting for gender, age, study year, student type, accommodation type, and religion, male gender (p = 0.007) and being international students (p = 0.021) were both significant correlates in the multivariate model. Compared to females, males were significantly more likely to engage in binge drinking, with the risk increasing by 195% (OR = 2.95). Similar to the drinking status model, international students showed a substantially lower risk of binge drinking, with the risk decreasing by 91% (OR = 0.09). As for the other four variables, none of them were significantly associated with binge drinking behavior (Table 3).

Table 4. Descriptive statistics and bivariate associations with number of standard drinks and AUDIT score.

Note: For outcome 3, n = 120, except for gender, in which case n = 119 due to 1 person being third gender. For outcome 4, n = 147, except for gender, in which case n = 146 due to 1 person being third gender. Superscript a refers to a correlation coefficient. Superscript b refers to an F statistic. Parental education and perceived family financial situation are both not statistically significant with respect to either outcome. ^*p < 0.05.

Descriptive statistics and bivariate associations for the number of standard drinks were calculated among students who reported alcohol use over the past year (n = 120). As shown in Table 4, in bivariate analyses, gender (t = −2.19,p = 0.032) was significantly associated with the number of standard drinks consumed on a typical day. Males reported higher alcohol consumption (Mean = 3.52, Std = 3.28) than females (Mean = 2.31, Std = 1.62). Besides, age (r = 0.22, p= 0.018) was also a significant correlate. There was no significant association between the number of standard drinks and the other seven sociodemographic correlates at the bivariate level.

In the multivariate model 1, gender (b = 0.272, p = 0.019) and age (b = 0.180, p = 0.003) remained significant correlates of the number of standard drinks after adjustment. Males consumed approximately 31.3% more standard drinks than females, while one additional year of age was associated with about a 19.7% higher number of standard drinks. Study year (b = −0.440, p = 0.038), which was not significant at the bivariate level, emerged as a significant correlate in the adjusted model. Compared with junior students, senior students consumed about 35.6% fewer standard drinks.

Model 2 added mother’s and father’s education, and perceived family financial situation, but the pattern of significant correlates remained largely similar. Gender (b = 0.321,p = 0.010) and age (b = 0.187, p = 0.004) remained significant, with only minor changes in effect sizes. However, study year (b = −0.421, p = 0.064) lost its significance in Model 2, and none of the parental SES correlates reached statistical significance. Although Model 2 showed a slight improvement in fit (R-squared = 0.163 vs. 0.133 in Model 1), there was a reduction in sample size (n = 109 in Model 2 vs. 119 in Model 1), due to missing data on mother’s and father’s education (Table 5).

For AUDIT score, study year (t = −2.27, p = 0.025), student type (F = 6.65, p = 0.002), and religion (t = −2.01, p = 0.046) were significantly associated with it in bivariate analyses, as shown in Table 4. There was no significant association between AUDIT score and the other six sociodemographic correlates at the bivariate level. Senior students (Mean = 4.96, std = 4.32) had a higher average score than junior students (Mean = 3.51, std = 3.81). The AUDIT scores of Mainland students (Mean = 4.60, std = 4.24) and local students (Mean = 3.95, std = 3.94) were both more than three times higher than those of international students (Mean = 1.25, std = 2.34). In addition, students with religion (Mean = 2.93, std = 3.76) had a lower average score than students without religion (Mean = 4.20, std = 4.05).

In the multivariate model 3, being international students (b = −1.045, p = 0.001) was the only correlate remaining significant. Compared to local students, international students had approximately 64.8% lower AUDIT scores. After adjusting for five other factors, gender (b = 0.309, p = 0.03) emerged as a new significant correlate, indicating that males had approximately 36.2% higher AUDIT scores than females. Study year (b = 0.126, p = 0.625) and religion (b = −0.060, p = 0.762) both lost their significance (Table 5).

Table 5. OLS regression models of number of standard drinks and AUDIT score.

Notes: n = 119 for Model 1, n = 109 for Model 2, n = 146 for Model 3, n = 131 for Model 4. ^*p < 0.05.

Model 4 also added three parental SES correlates. Gender (b = 0.342, p = 0.025) and being an international student (b = −1.162, p < 0.001) both remained significant and showed slightly larger effect sizes compared to Model 3. It was worthy of noting that accommodation type (b = −0.328, p = 0.045) became a significant correlate in Model 4. Compared to students living on-campus, students living off-campus had approximately 28.0% lower AUDIT scores. As in Model 2, mother’s and father’s education and perceived family financial situation all failed to reach statistical significance in Model 4. Besides, in Model 4, a similar improvement in fit was observed (R-squared = 0.197 vs. 0.158 in Model 3), and there was the same reduction in sample size (n = 131 in Model 4 vs. 146 in Model 3) for the same reason (Table 5).