Happiness In University Education*
Grace Chan, Paul W. Miller and MoonJoong Tcha
International Review of Economics Education, volume 4, issue 1 (2005), pp. 20-45
DOI: 10.1016/S1477-3880(15)30139-0 (Note that this link takes you to the Elsevier version of this paper)
The aim of this paper is to quantify the determinants of happiness in university students, with information drawn from a survey conducted with students at the University of Western Australia in 2003. An ordered probit model is applied. Happiness was linked to a range of factors, for instance, grades achieved, friendships developed, school facilities, opportunities to participate in extracurricular activities, and lecture quality. The findings reveal that the most important influences on the levels of satisfaction of students are school work, time management and relationships formed in university.
JEL Classification: A22
What makes university students happy? This question is important in all countries, though it has particular significance in Australia at the present time, as universities attempt to respond to government reforms that give them flexibility in the setting of fees for tertiary places. Already weight is placed in Australia on survey information that seeks to address aspects of this issue. The Good Universities Guide to Australian Universities, for example, includes survey information on the educational experience, which is based on a "within field" analysis of the Course Experience Questionnaire (CEQ) survey(note 1) of all coursework graduates, conducted yearly by the Graduate Careers Council of Australia. The CEQ survey gives an overview of the overall satisfaction of graduate students, provides detailed profiles of the teaching quality of each university and rates the universities according to how well students can acquire generic skills. Such survey information is intended to help students make their choice of university with the fullest possible information before them.
Nevertheless, today's students face difficulties when it comes to selecting a course and a campus that truly suits them. Under recent government reforms, Australian universities have the opportunity to increase fees, reduce government-funded places and increase full-fee paying places. Their capacity to do so will depend, in part at least, on students' satisfaction with their educational experience. This satisfaction could be linked to a range of factors, including grades achieved, friendships developed, school facilities, opportunities to participate in extracurricular activities, and lecture quality. There has not, however, been any systematic examination of these issues to date.
This study presents an account of the factors influencing satisfaction(note 2) among students at the University of Western Australia in 2003. The findings should contribute to the more general empirical research into the determinants of wellbeing (see Bradburn, 1969; Easterlin, 1974; Frey and Stutzer, 2002). They also have the potential to make three contributions to higher education management. First, university policy-makers may be able to use the results to identify the major determinants of student satisfaction, and thus be well positioned to develop a learning environment that will enhance students' university experience. For example, if there is a positive and significant relationship between participating in extracurricular activities and student satisfaction, school administrations could explicitly encourage or even expect student participation in such activities. Second, the research has the potential to provide new evidence on a range of topical issues concerning university life, including the roles of students' allowances, job income and grade achievements in influencing satisfaction. Third, the results obtained may allow students to organise themselves in order to attain their idea of the "good life".
The paper is organised as follows. The second section introduces the survey conducted and presents descriptive statistics on the overall levels of student satisfaction and on some of the simple correlations present in the data. The third section discusses results from an ordered probit model of the determinants of student satisfaction. The final section summarises the major empirical findings and implications of the results.
The Survey Data
Format of Survey
The data for this study were collected through a survey of around 1300 students attending core economics lectures at the University of Western Australia (UWA) in June 2003. The initial contact was made at the end of the lecture for the particular group, and participation was voluntary. Of the target sample, 931 students were enrolled in a first-year unit, 240 in a second-year unit, 81 in a third-year unit and 14 in a fourth-year (honours-level) unit.(note 3) The overall response rate was approximately 60 per cent, yielding 749 usable responses. Of these, 495 were from the first-year unit (response rate of 53.2%), 174 from the second-year unit (response rate of 72.5%), 66 from the third-year unit (response rate of 81.5%) and 14 from the fourth-year unit (response rate of 100%). The questionnaire(note 4) was designed to collect a wealth of information and was segmented into four parts. The first part covered the participants' demographic details, and comprised questions regarding their age, gender, method of entry, course that they are taking, year level, type of status (local or overseas student, full- or part-time student) and type of high school they attended. The succeeding parts included questions concerning the participants' employment history while at university, their life at university and their plans after university.
A Likert-type choice format was employed to measure the students' level of satisfaction. This scale has been widely used in prior research on well being/satisfaction (Bradburn, 1969; Easterlin, 1974; Di Tella, MacCulloch and Oswald 2001). The format of the present study allows more accurate response,(note 5) as subjects could indicate how much they agreed or disagreed with a series of statements on a 5-point scale ranging from 1 (strongly disagree) to 5 (strongly agree).
Basic Statistics of the Survey
Table 1 presents information on the gender and foreign and local composition of the students who enrolled in the core economics units and participated in the survey. Out of the 495 students from the participant pool enrolled in the first-year unit, 47.5% were males, 52.5% were females, 21.2% were foreign students and 78.8% were local students. The sample data on birthplace of students compares favourably with the unit enrolment data, though females are over-sampled. This relatively high response rate among females should not impact adversely on the analysis, given the apparent absence of gender effects in the statistical analysis.(note 6) Examination of the survey and unit characteristics for the other three units targeted reveals a broad correspondence. In particular, the sample proportions of females in these units do not differ significantly from the unit data. Also, a large proportion (74.8%) of the respondents came directly from high school, 6% from polytechnics and the rest through other entry modes (e.g. mature age entry). These statistics conform to what is known about the student population at UWA. Moreover, within the 749 usable responses, 48.1% of respondents were engaged in paid employment and worked an average of 11.2 hours weekly. This information is consistent with findings from a survey conducted by the Centre for the Study of Higher Education (CSHE) at the University of Melbourne in 1999, which showed that 51% of respondents had casual jobs and worked an average of 12.6 hours weekly.(note 7)
Table 1 Some basic statistics of respondents
|Unit Characteristics||Survey Characteristics|
|Year of Study||Males (%)||Females (%)||Foreign Students (%)||Local Students (%)||Males (%)||Females (%)||Foreign Students (%)||Local Students (%)|
The dependent variable for this study is an encompassing measure of student satisfaction (SA1) provided by responses to the question: "Overall, I am happy with my university life." A five-category Likert Scale was used, with respondents being asked to nominate one of five categories: 1=strongly disagree, 2=disagree, 3=neutral, 4=agree and 5=strongly agree. Information on the distribution of responses to the satisfaction question is presented in Figure 1.
Figure 1. Overall Satisfaction of Students
"Overall, I am happy with my university life"
Note: The number on the top of each bar is in percentages.
Referring to Figure 1, out of 749 responses, 13.9% of students strongly agreed that their lives at university were happy, 54.6% agreed that their lives in university were happy, while 1.2% strongly disagreed, 7.2% disagreed and 23.1% were neutral. The mean satisfaction level on the five-category Likert Scale employed is 3.73, and the standard deviation is 0.85.
In order to assess the reliability(note 8) of students' responses to the above question, a similar question was asked. Hence, students were asked to respond to the statement: "Overall, university life has been good for me." This satisfaction question (SA2) was placed towards the end of the questionnaire, whereas the first question (SA1) was placed towards the beginning of the questionnaire.
The results found for variable SA2 were similar to SA1, with the mean satisfaction level being 3.78, with a standard deviation of 0.82. Both satisfaction variables correlated significantly, with the correlation coefficient being 0.66. This correlation coefficient is often termed the "reliability" ratio. According to Ashenfelter and Krueger (1994), this coefficient implies that 66% of the content of SA1 is information and the balance is noise. As the questions were purposively designed to be similar rather than same,(note 9) the true reliability coefficient is likely to be understated by this correlation coefficient. Moreover, the cross-tabulation of SA1 by SA2 in Table 2 shows that most respondents to any given category of SA1 are in the same or an adjacent category of SA2. This information suggests that the satisfaction information is of an acceptable degree of reliability.
Table 2. Cross-Tabulation Table for Two Satisfaction Variables, SA1 and SA2
Preliminary examination of the data revealed a series of variables that seemingly have little relationship with student satisfaction and also a number of variables which seem to be quite important in this regard. Selected information is discussed below. Figure 2 presents information on the levels of satisfaction of males and females. The mean satisfaction for males and females are 3.77 and 3.70 respectively. 14.4% of the males and 13.5% of the females strongly agreed that they were "very happy in university life". The other categories of attitudes also reveal similar proportions of males and females.
Figure 2. Satisfaction of Students by Gender (Percentage of Respondents)
"Overall, I am happy with my university life"
A chi-squared test was used to examine independence of the responses of males and females to the satisfaction question. This generated a test-statistic of 4.07, which is less than the relevant critical value of 9.49. Thus the null hypothesis that male and female respondents do not differ in their (self-reported) satisfaction level cannot be rejected. This result is consistent with the findings of Frey and Stutzer (2000), who reported an absence of gender differences when they measured happiness using Swiss data, but it is inconsistent with the evidence of many other studies which show that females are more likely to be happy than males (e.g. Medley, 1980). The finding from the analysis of these student happiness data contrasts with the examination of academic performance among first-year students at the UWA by Win and Miller (2005), where females are shown to have average marks several points above those of comparable male students. The review, as well as the independent evidence, in Birch and Miller (2005) indicates that this finding relating to gender effects on academic performance carries across to other universities in Australia.
Figure 3 presents information on the level of satisfaction of students by age. The University of Western Australia enrolls both students who have recently completed high school and other students. This means there will be considerable variation in age within each unit. The mean age of participants is 19.48 (standard deviation of 3.1) and the range is 17-55. Within the large first-year unit, the mean age is 18.7 (standard deviation of 3.0) and the range is 17-55 while in the fourth-year unit the mean, standard deviation and range are 24.3, 3.9 and 20-32, respectively.(note 10) Hence, age effects may be related to year of study effects, and this issue is addressed in the statistical analysis below.
Figure 3. Satisfaction of Students by Age
Students who are 17-18 years of age recorded the highest mean level of satisfaction (3.82). As university students approach 19-20 years of age, the mean level of satisfaction drops to 3.67. Similarly, a trough occurs among the age group of 21-22, where the mean of satisfaction is 3.51. This U-shaped pattern with age among university students is consistent with the findings of Blanchflower and Oswald (2000).
The variation in the level of satisfaction with the age of students is likely to reflect interesting phenomena. New entrants' levels of happiness are likely to be influenced by the greater freedom at university compared with high school. Similarly, older students, who will be close to finishing their degrees,may be looking forward to the world of work, and their attitudes towards their university life may therefore be more positive. In between, students may find that higher levels of study progressively become both more academically challenging and involve greater volumes of work, each of which may be viewed as burdensome. Such perceptions might lower students' levels of satisfaction.
The environment in which students study is expected to influence their levels of happiness. Universities promote their campuses and facilities, including access to computers, books and academic help, interactive multi-media packages designed for certain disciplines and units, online discussions among students, virtual tutoring and obtaining lecture and tutorial materials online. The subjects of the survey were asked a series of questions pertaining to these types of resources and the general university environment, and the appraisal of each question is depicted in Table 3.
Table 3. An Appraisal of Resources and University Environment (percentage distribution)
|I am happy with my work environment (HWE)||13.77||59.63||18.45||6.95||1.2|
|The conditions of buildings and the sports facilities are good (GC)||14.04||56.02||22.26||6.95||0.73|
|I feel safe and secure (SS)||10.96||55.88||26.60||5.61||0.94|
Note: Numbers may not sum to 100 across rows due to rounding.
As shown in Table 3, 13.77% of respondents strongly agreed that they are happy with their work environment, while at the other extreme, only 1.2% strongly disagreed with the statement. Moreover, from the correlation matrix in Table 4, there exists a significant positive relationship between satisfaction and university environment. For instance, the correlation coefficient between the satisfaction variable, SA1, and variables, SS (I feel safe and secure), HWE (I am happy with my work environment), and GC (The conditions of buildings and the sports facilities are good) are significant, but only 0.24, 0.2 and 0.24 respectively.
Table 4. Correlation Matrix for Satisfaction and University Environment
Note: (*) Indicates the variable is significant at the 5% level. Refer to Table 3 for definitions of sub-variables.
Students' lifestyles presumably affect their reported levels of happiness.As time is scarce, students must learn how to manage their lifestyle properly, and mishandling of time may lead to stress and unhappiness. In the survey, the overwhelming majority of students reported "neutral" or "agree" for the following statements,(note 11) "I can balance work and university activities well" (BWUAW), "I can meet deadlines or goals in my university work" (MD) and "I have sufficient recreational and entertainment time outside the home" (SRT). Moreover, the correlation matrix in Table 5 shows that all the three variables were positively and significantly(note 12) correlated with the satisfaction variable (SA1). That is, as expected, students who managed time properly experienced more satisfaction.
Table 5. Correlation Matrix Between SA1, SRT, BWUAW, MD
Note: (*) Indicates the correlation is significant at the 5% level. SRT: I have sufficient recreational and entertainment time outside the home. BWUAW: I can balance work and university activities well. MD: I can meet deadlines or goals in my university work.
The brief review presented informs that across the satisfaction categories the variation as shown in Figure 1 appears to be related more to aspects of the university environment (Tables 3 and 4) and to students" lifestyles (Table 5) than it is to the students" personal characteristics (such as gender, examined in Figure 2). In the following section the links between student satisfaction and related factors are examined in the context of a multivariate model.
In this section, the data on students' levels of satisfaction are linked to a wide range of explanatory variables using an ordered probit model.(note 13) This model is particularly apt where the data are categorical and have an underlying ordering, and especially when the differences between categories of the dependent variables do not have the meaning of "distance" (Zavoina and McElvey, 1975). The basis of the ordered probit model is the linear relationship:
SAi = χ´iβ + εi
SAi is an unobserved index of satisfaction, and can be thought of as the underlying tendency of an observed phenomenon, namely the satisfaction scale SA1i. It is assumed that the random error, εi, follows a normal distribution. β is a vector of parameters, and χ is the vector of explanatory variables.(note 14) This vector includes age, gender, student monetary resources, extracurricular activities, satisfaction with school work, resources and the university environment, personal relationships formed, time management, health, whether employed in a part-time job, and perceptions of university reputation. These variables are described in full in the data. As the underlying tendency towards satisfaction among students, SA, is unobserved, the statistical analyses are based around the observed indicator, SA1, given by the responses to the statement "Overall, I am happy with my university life."
For the purpose of this analysis, these are classified as "strongly disagree (SA1=0)", "disagree (SA1=1)", "neutral (SA1=2)", "agree (SA1=3)", and "strongly agree (SA1=4)".
The observed scale, SA1, for individual i is linked to the unobserved variable, SA, as follows:
SA1i = 0 if SA1 ≤ μ0 (=0)
SA1i = 1 if μ0 ≤ SA1 < μ1
SA1i = 2 if μ1 ≤ SA1 < μ2
SA1i = 3 if μ2 ≤ SA1 < μ3
SA1i = 4 if μ3 ≤ SA1
where the μs are unknown threshold parameters separating the adjacent categories. These are to be estimated together with the ¦Â. The first threshold parameter is normalised to equal zero. With the normal distribution, the following probabilities may be calculated as:
Prob (SA1i= j) = Φ(μj - χ´iβ) - Φ(μj-1 - χ´iβ),
where Φ denotes the standardised cumulative normal distribution function.
Table 6 presents results for the ordered probit model outlined above. These estimates are computed for the sub-set of respondents who provided information on all the variables used in the estimating equations. Hence the sample size is 640, representing 85.4 percent of the initial sample.(note 15) The dependent variable is coded from 0 to 4. With the ordered probit model, a positive (negative) coefficient indicates a higher (lower) probability of membership of the top satisfaction category (of "strongly agree") and a lower (greater) probability of membership of the lowest satisfaction category (of "strongly disagree"). However, the effects of changes in any independent variable on predicted membership of the three intermediate satisfaction categories ("agree" , "neutral" , "disagree") are ambiguous. For this reason, the discussion of the estimated coefficients is kept general. More detailed analysis of the effects on the regressors on membership of the intermediate categories is presented later, with the aid of marginal effects and predicted probabilities. The chi-squared statistic for the ordered probit model is 284.65 and statistically significant, indicating that the joint test of all slope coefficients equaling zero is rejected. The Likelihood Ratio Index (or pseudo R2) for the ordered probit model is 0.188. This index gauges the extent to which the variables in the ordered probit model can explain satisfaction in university life and indicates model fit; in effect, it functions much like the coefficient of determination in the linear regression model (Greene, 2000). This pseudo R2 thus indicates a reasonable fit for the model.
Table 6. Satisfaction in University Life: Estimation by Ordered Probit
|Variables||Coefficients ("t" statistic)|
|Age 19-20||-0.105 (-0.95)|
|Age 21+||0.069 (-0.52)|
|Pocket Money and Job Income(a)||0.454 (2.05)**|
|Extracurricular Activities||0.096 (1.03)|
|Satisfaction with School Work||0.821 (9.52)***|
|Satisfaction with Resources and University Environment||0.163 (1.83)*|
|Time Management||0.267 (2.91)***|
|Good Health||-0.001 (-0.01)|
|Part-time Job||-0.126 (-1.32)|
|Degrees of Freedom||12|
|Likelihood Ratio Index (LRI)(b)||0.188|
|Prediction Success (%)(c)||60.63|
(***) Indicates the variable is significant at the 1% level.
(**) Indicates the variable is significant at the 5% level.
(*) Indicates the variable is significant at the 10% level.
μ1, μ2 and μ3 are threshold parameters.
a. The numbers should be multiplied by 10-3.
b. LRI= 1-LL ( )/ LL (0), where LL (b) is the value of the log likelihood at the estimated parameters and LL (0) is its value when all slope parameters are set to zero.
c. Prediction success compares the predicted actual distribution of the satisfaction data, where a student is assigned to a satisfaction category on the basis of the highest predicted probability of category membership.
d. The omitted categories for the binary variables are Age 17-18, male, does not engage in extracurricular activities, does not work. All other variables are continuous and are defined in the Appendix.
The threshold parameters in the Ordered Probit model,μ1, μ2 and μ3, are significant at the 1% level. Highly significant μ estimates indicate that the five satisfaction categories are distinct and hence favour using all five categories in an ordered probit model over aggregating some categories to form a dichotomous variable and using binary choice models.(note 16)
The significant explanatory variables that increase satisfaction levels in university are satisfaction with school work, good relationships formed, proper time management, good reputation of the university, income level and, to a lesser extent, satisfaction with the level of resources and the university environment.(note 17) The first four of these variables and the final one have similar metrics and hence the coefficients can be compared to assess relative degrees of influence on the probability index. This comparison shows that satisfaction with school work has the greatest impact on the probability index function (SA), followed by relationships formed and time management.
The explanatory variable "pocket money and job income" has a positive impact on the level of happiness, though the estimated impact appears small relative to that for the school work and time management variables. Hence, a rise in income by $175 (the mean weekly income) has an impact on the ordered probit index of less than one-tenth that of a one unit increase in the satisfaction with school work variable. While the sign of the coefficient on this variable is consistent with the prevalent thought that higher income equates with more happiness, the minor impact is disconcerting. However, Easterlin (1974) also found that income was a poor indicator of many measures of individual wellbeing. There may be many reasons why more money does not always translate into higher satisfaction.One commonly held viewpoint is that it is "relative" rather than "absolute" income that drives happiness (Blanchflower and Oswald, 2000; Easterlin, 1974). Another reason might be that students have to balance their time between job commitments, school work and leisure activities. As a result, earning extra income for themselves may not necessarily equate to increasing levels of satisfaction. In worse scenarios, working harder to afford more material goods can contribute to making people unhappier if they do not have sufficient spare time. A further possibility is that the effects of income on levels of happiness vary by age group: they are less important among the young age groups examined in this study, and more important among the older age groups that have tended to be the focus in prior research.
Note that the female binary variable is not significant in these analyses. In other words, the findings from the analysis of Figure 2 carry over to the multivariate analysis. The possibility that gender effects were being masked by an overly restrictive specification was explored by allowing the intercept and all the slope coefficients to vary by gender. However, the chi-squared test that all the slope coefficients and the intercept for females were, simultaneously, the same as those for males could not be rejected at the 5% level (χ2 test statistic of 18.72, which has a p value of 0.095). Similarly, the chi squared test that all the slope coefficients for females were, simultaneously, the same as those for males could not be rejected at the 5% level either.(note 18)
Finally, it is also noted that the age variables are not significant in the Table 6 results. As discussed earlier, it is possible that the age effects are related to the year of study. To examine this issue, year of study dummy variables were included in the estimating equation, along with the age variables. The year of study coefficients were numerically small and statistically insignificant. The point estimates on the ordered probit index were, from highest to lowest, for Year 2, Year 3, Year 1 and then Year 4. When the year of study variables were included in a model that did not contain the age variables, they remained statistically insignificant. Hence it appears that neither age nor year of study have much of an impact on students' levels of happiness.
The marginal effects from the ordered probit model are presented in Table 7. These effects are computed at the sample averages of variables. For the binary variables, they have been computed as the difference in the predicted distributions across the five categories for individuals possessing the characteristic associated with the binary variable, and the predicted distribution across the five categories for individuals who do not have the particular characteristic. In this instance, all variables other than the particular binary variable are set equal to the respective sample means. For each respondent, the marginal effect must be in the same direction as the average impact presented in the table for the "strongly agree" and "strongly disagree" categories, though for the intermediate categories the effects for specific individuals in the sample may vary even in sign from those presented. Nevertheless, the marginal impacts presented give a good indication of the overall changes in satisfaction levels as particular characteristics are altered.
The impacts of the significant explanatory variables are as follows. With a unit increase in satisfaction with school work, the probability of being classified as "strongly agree" to the statement "Overall, I am happy with my university life" increases by 12.31 percentage points and that of being classified as "strongly disagree" decreases by only 0.35 percentage points. Likewise, a unit increase in having good relationships formed in university increases the chance of being classified as "strongly agree" to the statement by 4.31 percentage points, and decreases that of being classified as "strongly disagree" by 0.12 percentage points. This result implies that forming good relationships with peers is good as satisfaction level in university will increase. Improvements in time management and in the perceived reputation of the university have impacts on the distribution across satisfaction levels that are similar to that described for relationships formed. The effect of a change in students' satisfaction with the resources available and the university environment on their distribution across the satisfaction categories is much more modest than that for these other variables. Finally, the effect of an increase in income on the distribution across the satisfaction categories is, as discussed previously, slight.
Table 7. Summary of Marginal Effects for Ordered Probit Models
|Pocket Money and Job Income**||0.01||0.01||-0.01||0.00||0.00|
|Satisfaction with School Work***||12.31||14.79||-20.82||-5.93||-0.35|
|Satisfaction with Resources and University Environment*||2.44||2.93||-4.13||-1.17||-0.07|
a. All numbers were multiplied by 100 for expositional purposes, thus the table is presented in the form of percentage point effects rather than effects on probabilities. Numbers may not sum to zero across rows due to rounding.
b. (***) Indicates the variable is significant at the 1% level.
(**) Indicates the variable is significant at the 5% level.
(*) Indicates the variable is significant at the 10% level.
c. (#) Indicates that it is a dummy variable.
The findings to date reveal that university students who are satisfied with their school work, possess good relationships with their peers, have proper time management, and who are happy with the reputation of the university and with the university environment report higher levels of satisfaction. While the marginal effects reported in Table 7 are revealing, additional insights into the links that these variables have with student satisfaction might be gained by computing predicted probabilities of membership of the various satisfaction categories for the range of values that the variables can take. The predicted probabilities have the advantage of informing on the levels of representation in the satisfaction categories, as opposed to changes in these levels. Hence, Table 8 presents the predicted distributions across response categories to the statement "Overall, I am happy with my university life" derived from the ordered probit model for the aggregated school work and time management variables. The predicted probabilities are specified for up to five values for the respective variables, with all other explanatory variables held at their mean values.
Table 8. Predicted Probabilities at the Specified Values
Source: Ordered probit estimates, Table 6. Numbers may not sum to 100 across rows due to rounding.
Both school work and time management variables have immense effects on the distribution of happiness. For the school work variable, which has a mean value of 3.4, the predicted probability for strongly agreeing to the statement at value 1 is less than 1% and this rises to 46.5% at value 5. Correspondingly, the percentage strongly disagreeing to the statement falls from 14.9% at value 1 to zero at value 5. The representation in the agree category is 8.72% at the lowest value for the satisfaction with school work variable, and 50.78% at the highest value. While there are reasonable representations in each of the "neutral" and "disagree" categories at value 1 of the satisfaction with school work variable, very few students are in these categories if they are highly satisfied with their school work. Likewise, with regard to the time management variable, the predicted probability for strongly agreeing to the statement at value 5 is 14.55%, some 13 percentage points higher than at value 1. In this instance, higher values of time management are shown to be largely associated with reductions in the probability of being in the neutral and disagree categories. Hence, Table 8 points conclusively in the direction that both explanatory variables drive satisfaction substantially.
The two influences relating to school work and time management investigated in Table 8 are captured in the model using composite variables, created by aggregating the categorical responses to a number of components. The five components that make up the highly significant composite variable school work were derived from responses to the statements (mnemonics are in parentheses): (i) "I am given the chance to do work that really interests me" (DWTI); (ii) "I am happy with the marks I have achieved so far in university" (HWM); (iii) "I can cope with my university work well" (COPE); (iv) "I enjoy studying my course subjects" (ES); (v) "I achieve a standard in my work which I consider satisfactory" (AAS). Correspondingly, the other highly significant composite variable, time management, encompasses responses to three statements: (i) "I can balance work and university activities well" (BWUAW); (ii) "I can meet deadlines or goals in my university work" (MD); (iii) "I have sufficient recreational and entertainment time outside the home" (SRT). This section aims to ascertain whether all of these more fundamental components influence satisfaction in the same way.
Table 9 presents results from an ordered probit model where the school work and time management variables are replaced by the sub-variables mentioned above. This alternative model is considered in order that the sources of the school work and time management effects might be identified. Numerically, it was found that those effective in coping should have greater satisfaction, ascertained by both the sizes of the partial effect on the ordered probit index and the t-statistic, of 0.366 and 5.21, respectively. Other important influences include students being given the chance to do work that interests them, students enjoying studying their course subjects, and students being happy with the marks they have achieved. Remarkably, achieving a satisfactory standard in their coursework is not a statistically significant contributor to students' overall level of satisfaction.
Table 9. Satisfaction in University Life: Ordered Probit Model (With Sub-Variables)
|Variables||Coefficients ("t" statistic)|
|Age 19-20||-0.098 (-0.89)|
|Age 21+||-0.072 (-0.54)|
|Pocket Money and Job Income||0.479b (2.17)b**|
|Extracurricular Activities||0.090 (0.96)|
|AAS (School Work)||0.010 (0.15)|
|HWM (School Work)||0.146 (2.36)**|
|ES (School Work)||0.199 (2.93)**|
|DWTI (School Work)||0.184 (2.99)***|
|COPE (School Work)||0.366 (5.21)***|
|Satisfaction with Resources and University Environment||0.157 (1.74)*|
|Relationships Formed||0.303 (4.82)***|
|BWUAW (Time Management)||0.008 (0.11)|
|MD (Time Management)||-0.042 (0.60)|
|SRT (Time Management)||0.210 (4.22)***|
|Good Health||-0.010 (-0.14)|
|Part-time Job||-0.123 (1.28)|
|Degrees of Freedom||18|
|Likelihood Ratio Index (LRI)c||0.200|
|Prediction Success (%)d||59.38|
For notes to table, see Table 6.
While the estimates in Table 6 showed that proper time management was important to general levels of satisfaction in university life, the detailed analyses conducted here reveal that the only significant component of the time management variable is SRT. This implies that students who have sufficient recreational and entertainment time are happy. On the contrary, the level of satisfaction among students who "can balance work and activities well" and "can meet deadlines or goals" does not differ significantly from that of students who cannot perform these functions.
The effect of these two explanatory variables, school work and proper time management, on student satisfaction can also be explained by the set of expectations students have concerning how much work and time they think is demanded of them. Students expecting favourable outcomes in their lives work hard to achieve the goals they have set for themselves. In contrast, students expecting failure tend to disengage from moving towards their goals. If they do not foresee this reasonably accurately, students mishandle time, and this may result in dissatisfaction. Forming good relationships with peers is another vital component constituting satisfaction in university life.
Finally, it is noted that the diversity of the impacts of the components used in the construction of the composite school work and time management variables may call into question the practice of using such composite variables. Constructs along these lines are widely used in the economics of education literature (e.g., Le and Miller, 2004; Marks, Fleming, Long and McMillan, 2000). In the current analysis, each composite variable has been formed as the average of its underlying components, there being three such components for the time management variable and five for the school work variable. Hence the use of these composite variables could be viewed as being valid if each of the components had a coefficient that was onethird of the size of the time management coefficient in Table 6, and one-fifth of the size of the school work coefficient in Table 6. Tests of these restrictions revealed that each could be rejected at the 5% level of significance (but not at the 1% level).
Hence, researchers should consider carefully the trade-off between the parsimonious specification that can be obtained through the use of composite variables and the possibly unjustified restrictions inherent in such an approach.
This paper investigates the roles that economic and social factors play in determining the happiness levels experienced by university students. As well as contributing to understanding happiness, the findings also have important implications for the design of university curriculum. The most important influences on the levels of satisfaction of students are school work, time management and relationships formed in university. This suggests that if the enhancement of student satisfaction is a priority, policy makers should examine student workloads. In addition, they can encourage courses that aid students with better time management and promote forging of good peer relationships in university.
While this study provides interesting findings, it is important to recognise the limitations of the research. The most important of these are that the data are restricted to one university and to students attending one particular sequence of units. This limitation points to the direction that further research could take. One possible line of research would be to include all Australian universities with larger sample sizes, and conduct the surveys at different time periods, so that details of the individual differences in the students can be more fully explored and any regional difference can be found. This work should also extend the coverage of students so that subjects other than economics are included. Moreover, it would be interesting to expand the scope of the research by linking the decisions that students take (for example, dropping out, switching courses, and the pursuit of further studies) to their reported levels of satisfaction. A further way the scope of the research could be expanded is for information on academic performance to be collected, perhaps through linking of student responses to surveys of this type to institutional data (covering the academic basis for admission to university, grades achieved at university). The research could also be extended to other countries, particularly with a view to ascertaining cultural influences on happiness levels among university students.
These expansions should result in a better understanding of happiness in education. Although there is room for methodological concerns and apprehension over the quality of happiness data, economists should, however, not be too critical of what they measure and use. The main use of a happiness measure is to identify the determinants of happiness rather than to compare levels of subjective wellbeing. What matters are finding out what type of happiness data is apt for answering particular questions and for addressing particular issues.
Birch, E. R. and Miller, P. W. (2005). "The Determinants of Students' Tertiary Academic Success" , in J. Pincus (ed). Microeconometric Tools for Policy Research, Productivity Commission (forthcoming).
Easterlin, R. A. (1974). "Does Economic Growth Improve the Human Lot? Some Empirical Evidence" in David, P. A. and Reder, M. W. (eds.), Nations and Households in Economic Growth: Essays in Honor of Moses Abramovitz, New York and London: Academic Press, pp. 89-125.
Marks, G. N., Fleming, N., Long, M. and McMillan, J., (2000), Patterns of Participation in Year 12 and Higher Education in Australia: Trends and Issues. LSAY Research Report Number 17, Australian Council for Educational Research, Victoria.
The variables used in this study are defined as follows.
Satisfaction in University (SA1): The respondents' overall feelings towards their university life whereby the responses to the statement: "Overall, I am happy with my university life" were recorded in one of five categories: "strongly agree" , "agree" , "neutral" , "disagree" and "strongly disagree" .The dependent variable formed from this information was defined as 0 if the individual strongly disagreed, 1 if disagreed, 2 if neutral, 3 if agreed, and 4 if strongly agreed.
Age17-18: This is a binary variable, set equal to one where the student is 17 or 18 years old, and to zero otherwise.This is the omitted (or benchmark) category in the analysis.
Age19-20: This is a binary variable, set equal to one where the student is 19 or 20 years old, and to zero otherwise.
Age21+: This is a binary variable, set equal to one where the student is 21 years old and above, and to zero otherwise.
Gender: This is a dichotomous variable and is set to unity if the individual is female.
Pocket Money and Job Income: This is a continuous variable and measures the total amount of money a student receives from parents, the government and parttime jobs.
Extracurricular Activities: This is a dichotomous variable and is set to unity if the student partakes in university extracurricular activities.
Satisfaction with School Work: A measure of each student's general satisfaction with school work, which involves doing work that interests the student, obtaining satisfactory results, coping with and enjoying university work. A continuous variable was formed by as the mean of the categorical responses concerning school work for each student.
Resources and University Environment: This variable is created from information obtained on student's satisfaction with the university environment and access to resources. It covered whether students were happy with their work environment, whether conditions of buildings and sports facilities were good and whether they felt safe and secure. A continuous variable was formed by aggregating the categorical responses pertaining to these issues for each student.
Relationships Formed in University: The basis of this measure is the student's development of good relationships with school mates. A continuous variable was formed from the categorical responses relating to this issue for each student. Time Management: Having sufficient recreational and entertainment time outside the home, balancing work and university activities well and being able to meet deadlines or goals in school work form an indicator of time management. A continuous variable was formed from as the mean of the categorical responses concerning these issues for each student.
Health: This is a continuous variable and measures the student's self-reported good health conditions.
Part-time Job: This is a dichotomous variable and is set to unity if the student has a part-time job; otherwise it is defined to zero.
Reputation: This variable is created from information obtained on student's satisfaction with the reputation of the university they are studying in. A continuous variable was formed as the mean of the categorical responses concerning these issues for each student.
The small number of students who failed to report answers to any of the questions used in the construction of these variables were excluded from the analysis.The means and standard deviations of the main variables are presented below.
|Satisfaction in University Life||3.731||0.826|
|Satisfaction With School Work||3.402||0.643|
|Satisfaction with Resources and University Environment||3.747||0.575|
|Relationships Formed at University||3.875||0.791|
|Achieve a Standard of Work Considered Satisfactory (AAS)||3.459||0.864|
|Happy with Marks (HWM)||3.139||0.992|
|Enjoy Studying (ES)||3.470||0.850|
|Do Work that Really Interests (DWTI)||3.370||0.914|
|Cope with University Work Well (COPE)||3.355||0.853|
|Balance Work and University Activities Well (BWUAW)||3.491||0.799|
|Can Meet Deadlines (MD)||3.898||0.723|
|Sufficient Recreational and Entertainment Time (SRT)||3.747||1.011|
We are grateful to students at the University of Western Australia for participating in the survey upon which this paper is based, and to two anonymous referees for many helpful comments. Miller acknowledges financial assistance from the Australian Research Council. Opinions expressed in this paper are those of the authors, and should not be attributed to the funding agency or to the University of Western Australia.
 The terms satisfaction and happiness are often used interchangeably in the recent literature and are used interchangeably in this paper.
 The large first-year unit had two streams: the other three units were taught in a single class.
 The questionnaire was similar in construction to those used in the Longitudinal Surveys of Australian Youth. The questionnaire, and an EXCEL file containing the data, are available upon request from the corresponding author.
 Given the absence of gender effects, attempts were not made to generate sample weights by gender and year level that could account for the overrepresentation of females, particularly in the first-year unit.
 As the survey was quite short (respondents took fewer than ten minutes to complete it), it was considered prudent to ask a similar question rather than the same question.
 For the second year unit, the mean, standard deviation and range are 20.4, 2.1 and 18-29, respectively, and for the third-year unit they are 22.1, 3.0 and 17-33, respectively.
 The short form version of the time management variables are in parentheses.
 The t-statistics for BWUAW, SRT, MD are 6.94, 4.01, and 3.88.
 The model was also estimated using OLS. The findings from this alternative method of estimation were similar to the ordered probit results reported in Table 6.
 The unknown parameter vector in the empirical model was estimated by using LIMDEP version 8 (see Greene, 2002).
 Most missing values were due to income data not being available (89 out of 109). Note that respondents reporting zero weekly income are included in the analysis. Approximately 14% of the sample are in this category, which corresponds with information from the 2001 Australian Census of Population and Housing on income levels among full-time students aged 20-24 years (14% report zero income), and even among both part-time and full-time students in this age category (11% report zero income).
 As the first category ("Strongly Disagree") has a small membership, the model was also estimated where this category was combined with the adjacent category (of "Disagree"). The only change to the results was that the variable for satisfaction with resources and the university environment, significant at the 10% level in Table 6, became insignificant (p value of .107).
 An examination using the ordered probit model of the impacts of school type variables such as students coming from Catholic, Independent, Government and Polytechnic schools, along with other variables such as the mean mark on assessments and whether students were local or overseas, economics or of other degree-types, revealed that there were statistically insignificant. Thus, these variables were omitted from the regression. The insignificant regressors which were retained in the model are for variables that are typically found in analyses of the determinants of happiness. There did not appear to be any appreciable gain in efficiency from omitting these regressors.
 χ2test statistic of 17.56, which has a p value of 0.092. Similarly, extending the test of gender differences to consider the thresholds in the ordered probit model as well did not uncover a statistically significant effect.
University of Western Australia
Crawley, WA 6009, Australia
Tel: (618) 6488 2780
Paul W. Miller
University of Western Australia
Crawley, WA 6009, Australia
Tel: (618) 6488 2924
University of Western Australia
Crawley, WA 6009, Australia
Tel: (618) 6488 2780