Education is widely accepted as a learning instrument for promoting economic growth and development. Through education people can obtain requisite know-ledge, skills and expertise which increase their pro-ductive capacity thereby enhancing economic growth and development. Students are most essential asset for any educational institute. Students academic per-formance plays an important role in producing the best quality graduates who will become great leader and manpower for the country thus responsible for the countrys economic and social development. Students who have access to quality educational programs per-form better and successful in their lives. In a develop-ping country like Bangladesh education gains even more importance. But in developing countries, the problem of poor academic performance leads to the widely acclaimed fallen standard of education. Con-sidering the critical role of education in nation building and national development, countries particularly deve-loping economies have over the years continue to in-troduce policy reforms in the education sector. By identifying the socio-economic factors having signi-ficant impact on students academic performance in higher education, government will be able to design and implement the policies to improve the students academic performance and the quality of education. Different researchers researched on various variables and a lot of different variables have been studied.  All of the previous research reviews support the hypo-thesis that students academic performance depends on different socio-economic, psychological and environ-mental factors. Anderson et al. (1994) performed a study on the determinants of success in university. The study shows that the result of the secondary school is consistently the best indicator of the result of college students. Also same study conveyed by them at uni-versity level. The estimated result of the study shows that previous academic background has an important positive impact on undergraduate students intro-ductory economics courses. 

Another study conveyed by Considine and Zappala, (2002) shows that parents providing higher level of psychological support for their children through envi-ronments encourage the betterment of skill important for success at school. The study also implies that low socio - economic status of families negatively affect childrens academic achievement because low socio-economic status prevents access to necessary resources and creates additional stress at home. Nayebzadeh et al. (2011) performed a study on the impact of back-ground variables on educational performance in Zim-babwe using primary data for secondary and tertiary level of education. They performed a comparative study on the academic performance for both male and female students and showed that academic perfor-mance of female students are better than male students. They further explained that age, environment and access to internet are some of the important variables of academic achievement of undergraduate students. Cheesman et al. (2006) conveyed a research on deter-minates of students academic attainment at graduation level reflection from the Caribbean on 900 sample students. The study indicates that gender gap in favor of male students is only related with university ad-mission exam scores. They pointed out the reason that female students perform or score low grade in admi-ssion exam result and they joined in less competitive departments and this situation enables female students performing better or score high CGPA in undergrad- duate level than male students (Rashid & Islam, 2020). 

Aldin et al. (2011) conducted another study on the determinant of academic performance of under-graduate students. They tried to find out the rela-tionship among different variables related with per-sonal and family backgrounds and the average mark of students. They found that socio-economic background of family and good personal background of students contribute to better academic achievement. Another study performed by Bloom et al. (2006) on under-graduate students academic performance of econo-mics in Osun state. The study shows that socio-eco-nomic structure of family has an important impact on academic performance of undergraduate students. According to this study, family income, education level of parents, students previous academic results and university admission exam score of students are important factors having significant impact on students academic achievement. All these factors have positive impact on students academic achievement. Martha, (2009) conveyed a study on factors affecting academic performance of undergraduate students at Uganda Christian University. The findings of the study display that parents economic status, university ad-mission test score and students regular class atten-dance all are have positive impact on undergraduate students academic performance (Cada, 2021). The estimated slope coefficient of parents economic sta-tus, university admission test score and class atten-dance are statistically significant. Weldegiorgis and Awel, (2013) conducted a study on the determinants of students attrition at Mekell University. The econo-metrics results of the study show that students gender, study hour per day, national entrance examination overall results, mathematics result, students alcohol drinking behavior, fathers education level, mothers education level and family financial constraint all are important determinants of students academic per-formance. Study hour per day, national entrance exa-mination overall results, mathematics result, fathers education level and mothers education level have significant positive impact on students academic per-formance. And students gender, age, students alcohol drinking behavior and financial constraint have significant negative impact on students academic performance. 

Mersha et al. (2013) performed a research on factors affecting female students academic achievement, using a sample of 600 students on second year and above undergraduate female students at Bahir Dar University. They found that academic achievement of female students affected by students personal related factors, university related factors and economic related factors. Moreover, previous academic background is an important factor for their low academic achieve-ment. Another study performed by (Ali et al., 2013) on factors contributing to the academic performance of graduate students of University of Bahawalpur Rahim Yar Khan Campus. The study shows that they took student academic performance (Test Score) as a depen-dent variable and age, gender, study hour, guardian schooling year, guardian economic status/income, admission test score, residential area, medium of school, tuition and accommodation as independent variables. The results of the study display that age, study hour and income have statistically significant impact on graduate students academic performance. The impact of other factors may also have impact on academic performance but it is not statistically signi-ficant in their study.  Hijazi and Naqvi, (2006) per-formed a study on factors affecting students academic performance. They found that there is a significant positive relationship between previous academic results and students academic performance. More-over, class attendance and mothers education have positive impact on students academic performance.  The study of Yigermal, (2017) displays the presence of significant relationship between gender differences, university entrance exam, studying hours, students former academic back ground, students behavior on taking of alcoholic drug and academic performance (CGPA). Raychaudhuri et al. (2010) conducted a study on the socio-economic factors affecting students academic achievement. The key findings of the study indicate that socio-economic factors like class atten-dance, family income, education level of father and mother, teacher-student ratio, presence of trained teacher in school, sex of student and distance of school are important factors that affect academic performance of students. The research study of Gustafsson et al. (2018) displays that many factors such as socio-economic status of family, learning facilities; gender, age differences, and previous academic background have significant impact on students academic perfor-mance. Saenz et al. (1999) performed a study on the determinants of students academic performances. The findings of the study show that the academic perfor-mances of students are positively and significantly related to the use of library and the level of their parental education. 

Another study on the factors affecting students academic performance conducted by Mushtag and Khan, shows that students face a lot of problems in developing positive study attitudes and study habits. Guidance is an important factor that can improve students study attitudes and study habits. The proper guidance from parents and teachers indirectly affects the performance of the students. The study of Kirmani & Siddiquah, (2008) implies that the academic envi-ronment, fathers education and previous results are important variables that affect academic performance. Noble et al. (2006) showed the significant relationship between children academic achievement and family background variables. Another study performed by Shabani, (2013 shows that age, marital status, resi-dence status, class size, occupation of parents and pre-vious academic performance are important determi-nants of performance in examination. The main objective of this study is to investigate the impact of different socio-economic factors that affect under-graduate students academic performance. 

In this article, different socio-economic variables like gender, age, class attendance, study hour, university admission test score, previous academic results, family income, fathers education, mothers education, part time job, participation of extra-curriculum activities and group study are used to find out the important determinants of undergraduate students academic achievement or CGPA (Cumulative grade point aver-age). This research will be different from previous researches in a way that it is the first research in which some independent dummy variables like students part time job participation, internet using for education pur-pose, students participation in extra-curriculum acti-vities and group study will be studied. Recently Bangladesh has achieved the status of lower middle income country and GDP growth rate has reached more than 8 percent. In order to maintain this econo-mic growth and development, Bangladesh has to give more focus on the improvement of the quality of education and human development. As a developing country, Bangladesh needs to utilize the resources pro-perly the country has. Therefore, the outcome of this study will be helpful for both parents and government concerned authority to take proper steps in order to improve overall educational attainment in Bangladesh.

In this study, an Analysis of covariance model (AN-COVA) is applied to find out the different socio-economic factors undergraduate students academic results or CGPA. An ANCOVA model is the model in which dependent variable is the function of quantit-ative variables as well as dummy variables (Guajarati, 2004). Dummy variable is the variable in which it takes value 0 and 1. In such type of variables, 0 indi-cates absence of an attribute and 1 indicates presence of an attribute. In this paper ANCOVA model is con-structed by different types of socio-economic variables where  undergraduate students academic performance or CGPA (Cumulative grade point average ) is depen-dent variable and class attendance, per day study hour, previous academic results, gender, family income, fathers education, mothers education, students parti-cipation of part time job, participation of extra curri-culum activities, participation of group study and using internet for education purpose  these are independent variables.

Econometric model - The basic ANCOVA linear model can be written as:

CGPAi = α1 +β1Gi + β2CAi + β3SHi + β4PRi+ β5ATSi + β6FIi + β7FEi + β8MEi + β9PJi + β10IUi + β11EAi + β12GSi + Ui ….…………………………………………. (1)

In the above equation (Authors compilation),

To apply OLS (Ordinary Least Square) method on this model and ensure the estimated slope coefficients are BLUE (Best Linear Unbiased Estimator), the above model is assumed as linear in parameters, with zero mean value of error term and no autocorrelation bet-ween error terms.

Sample and sampling technique - A cross- sectional research design allows the collection of data in a particular region at one point in time. The cross -sectional design is the most efficient method of data collection in terms of time and financial resources (Kothari, 2004). In this study, a cross-sectional res-earch design is applied to collect primary data from 300 randomly selected undergraduate students of Social Science faculty, University of Dhaka. Among them, 164 are male students and 136 are female students. Sample students have been selected randomly through simple random sampling method and a struct-ured questionnaire containing both closed and open ended questions has been used to collect data. Both quantitative and qualitative primary data have been used in this study. The selection of sample size is based on statistical principles. In this study, each department is considered as a stratum and the samples have been selected from each stratum. Both probability and non- probability sampling techniques are applied to select samples. In this study, Slovins formula is applied in calculating the sample size of 300 under-graduate students. The Slovins formula is as follows (Dhokhikah et al., 2005).

η =N/(N*d^2+1)

Where, η = Number of sample, N = Total number of population, and d = Error margin.

Data - This paper is based on direct sample survey of 300 students among them 164 are male students and 136 are female students. The statistical properties of sample data are demonstrated below:

Table 1: Statistical Properties of Data Set (Source: Authors calculation).

The above statistical table shows that the mean CGPA (Cumulative grade point average) is 3.25 while mini-mum and maximum CGPA are 2.18 and 4 respect-ively. Approximately, 59 percent students including both male and female students have at least 80% class attendance. And the remaining 41 percent students have less than 80% class attendance. Per day mean study hour for each student is 3.09 hour, where mini-mum and maximum per day study hour of students are 1 and 9 hour respectively. The mean value of previous results which is the average GPA (Grade point aver-age) of SSC and HSC results for each student is 4.63. The maximum previous average GPA is 5 and mini-mum is 3.5. The mean value of students university admission test score is 68.55 where maximum and minimum admission test score are 51 and 89.15 out of 120.The average monthly family income of respon-dents students is 31.81 thousand BDT. Where the maximum and minimum monthly incomes are 120 and 8 thousand BDT. The average year of the school-ing of fathers of the respondent students is 11.62 years and the average year of the schooling of mothers is 9.93 years. Among 300 students, 67.67% students have performed part time job like tuition, coaching etc. in their graduation period. Moreover, 65.67 % students have used internet facilities for their education pur-pose. And 63% students have participated in group study for education purpose in their graduation period. Among 300 students, 201 students that is 67% students have participated in different extra-curriculum acti-vities. 

Regression Analysis - The result of the regression analysis is given below:

Table 2: Result and hypothesis test (Source: Authors calculation).

Estimated equation is given below:

(CGPA) ̂i = 1.1233 -0.0036Gi+0.2182 CAi + 0.0522 SHi + 0.1389 PRi + 0.0111 ATSi + 0.0025 FIi + 0.0163 FEi + 0.0134MEi – 0.0607 PJi + 0.0079 IUi -0.0100EAi + 0.1241 GSi ……………………………………….. (2)

The above estimated equation shows the relationship between dependent and independent variables. In the above estimated equation, average CGPA (Cumulative grade point average) varies 0.0036 (Out of 4.00) across gender of students with reference category of dummy variable is female. Ceteris paribus, average CGPA is lower by on average 0.003 for male students compare to female students. That means, female students per-forms better than male students in their academic examinations. Since the P- value (The exact level of significance at which null hypothesis can be rejected; N Gujarati) of the estimated slope coefficient of the dummy variable gender is 0.921 which is greater than 0.05. So we cant reject null hypothesis on the basis of t- test at 5% level of significance. That is, the esti-mated slope coefficient of the dummy variable gender is not statistically significant from zero. Here, the estimated slope coefficient of dummy variable CA (Class attendance) is 0.2182. Ceteris paribus, CGPA increases on average by 0.2182 for those students who attend at least 80% of classes. The P- value of the esti-mated slope coefficient of dummy variable CA (class attendance) is 0.000 and absolute t- value is 5.34. So, we can reject null hypothesis on the basis of t- test at any level of significance. That means, class attendance has a significant impact on CGPA. Again the value of estimated slope coefficient of explanatory variable SH (Study hour) is 0.0522. Ceteris paribus, if the per day study hour of students increases by an hour, it will increase CGPA on average by 0.0522.The P- value of the estimated slope coefficient of study hour is 0.000 and absolute t- value is 4.47. So, we can reject null hypothesis on the basis of t-test which is used for testing individual slope coefficient. In other words, the estimated slope coefficient of SH (Study hour) is statistically significant different from zero. The value of the estimated slope coefficient of PR (Previous results) which is the sum of the average of SSC (Secon- dary school certificate) and HSC (Higher secondary certificate) GPA (Grade point average) is 0.1389. Ceteris paribus, if the average GPA of previous results increases by 1.00 point out of 5.00, it will increase CGPA on average by 0.1389. The P- value of the estimated slope coefficient of previous results is 0.003 and absolute t- value is 2.99. So, we can reject null hypothesis that previous results have no significant impact on CGPA of undergraduate students. That means, estimated slope coefficient of previous results is statistically significant different from zero. The value of the estimated slope coefficient of university admission test score is 0.0111. There is a positive relationship between university entrance examination score and graduation level CGPA of sample students. Ceteris paribus, average CGPA of a student increases on average by 0.0111 if that particular student gets one mark more in his university admission test. The P- value of the estimated slope coefficient of ATS (Admission test score) is 0.000 and absolute t- value is 3.84.So we can reject null hypothesis that university admission test score has no significant impact on graduation level CGPA on the basis of individual slope coefficient t-test. So the estimated slope coefficient of ATS is statistically significant from zero at any level of significance which means that university admission test score has a significant impact on undergraduate students academic performance. The value of the esti-mated slope coefficient of FI (Family income) is 0.0025. Ceteris paribus, if average monthly family income increases by one thousand BDT (Bangladeshi taka), it will increase CGPA on average by 0.0025. The P-value of the estimated slope coefficient of monthly family income is 0.041 and absolute t – value is 2.05. So, we can reject null hypothesis that monthly family income has no significant impact on CGPA on the basis of t- test. The estimated slope coefficient of monthly family income is statistically significant different from zero at 5% level of significance. 

The value of the estimated slope coefficient of FE (Fathers education) is 0.0163. Ceteris paribus, if complete years of fathers education increases by one year, it will increase CGPA on average by 0.0163. The P-value of the estimated slope coefficient of FE (Fat-hers education) is 0.015 and absolute t- value is 2.44. So we can reject null hypothesis that FE (Fathers education) has no significant impact on CGPA on the basis of t- test. So, the estimated slope coefficient of FE (Fathers education) is statistically significant different from zero at 5% level of significance. The value of the estimated slope coefficient of ME (Mot-hers education) is 0.0134. Ceteris paribus, if complete years of mothers education increases by one year, it will increase CGPA on average by 0.0134. The P-value of the estimated slope coefficient of ME (Mot-hers education) is 0.029 and absolute t- value is 2.20. So we can reject null hypothesis that ME (Mothers education) has no significant impact on CGPA on the basis of t- test. So, the estimated slope coefficient of ME (Mothers education) is statistically significant different from zero at 5% level of significance. The value of the estimated slope coefficient of dummy variable PJ (Part-time job participation) with reference category of not participation in part time job is -0.0607. Ceteris paribus, if a student participates in part time job during his/her graduation period, it will decrease his/her CGPA on average by 0.0607. The P-value of the estimated slope coefficient of PJ (Part-time job participation) is 0.102 and absolute t- value is 1.64. So we cant reject null hypothesis that PJ (Part-time job participation) has no significant impact on CGPA on the basis of t-test. So, the estimated slope coefficient of PJ (Part-time job participation) is not statistically significant different from zero at 5% level of significance. 

The value of the estimated slope co-efficient of dummy variable IU (Internet using) with reference category of those students who do not use internet for education purpose is 0.0079. Ceteris paribus, CGPA on average increases by 0.0079 for those students who use internet for their education purpose. The P-value of the estimated slope coefficient of IU (Internet using) is 0.854 and absolute t- value is 0.18. So we cant reject null hypothesis that IU (Internet using) has no signi-ficant impact on CGPA on the basis of t-test. So, the estimated slope coefficient of IU (Internet using) is not statistically significant different from zero at 5% level of significance. The value of the estimated slope coefficient of dummy variable EA (Extra-curriculum activities) with reference category of those students who do not participate in extra-curriculum activities is -0.0100. Ceteris paribus, if a student participates in extra-curriculum activities during his/her graduation period, it will decrease his/her CGPA on average by 0.01 The P-value of the estimated slope coefficient of EA (Extra curriculum activities participation) is 0.849 and absolute t- value is 0.19. So we cant reject null hypothesis that EA (Extra curriculum activities) has no significant impact on CGPA on the basis of t-test. So, the estimated slope coefficient of EA (Extra curri-culum activities) is not statistically significant different from zero at 5% level of significance. 

The value of the estimated slope co-efficient of dum-my variable GS (Group study) with reference category of those students who do not parti-cipate in group study for education purpose is 0.1241. Ceteris paribus, CGPA on average increases by 0.1241 for those students who participate in group study for education purpose. The P-value of the estimated slope coefficient of GS (Group study) is 0.008 and absolute t- value is 2.66. So, we can reject null hypothesis that GS (Group study) has no significant impact on CGPA on the basis of t-test. So, the estimated slope coeffi-cient of GS (Group study) is statistically significant different from zero at 5% level of significance. The value of intercept term in this estimated regression equation is 1.1233. 

The P- value and absolute t- value of the intercept term are 0.00 and 4.59 respectively. So, this intercept term is statistically significant on the basis of t-test. That indicates the logic behind the inclusion of intercept term in this regression equation. The signs of all the slope coefficients of the above estimated regression equation is relevant to our priori expectations.

Test of Goodness of Fit and ANOVA Table - The value of R-square is commonly known as the co-efficient of determination and most commonly used to measure of the goodness of fit of a regression line. The value of R-square measures the proportion or percent-tage of the total variation in dependent variable explained by the explanatory variables or regression model. The value of adjusted R- square also measures the total variation in dependent variable due to ex-planatory variables adjusted with the degree of free-dom. 

The value of R-square and adjusted R-square of this regression equation are 68.04 percent and 66.71 percent respectively. So in the above regression model, the dependent variable is well explained by the explanatory variables. The overall significance of a multiple regression test is performed by using F-test. The result of the ANOVA (Analysis of Variance) of this regression model is given below:

Table 3: ANOVA Test Results (Source: Authors calculation).

In the above ANOVA model, F value is 50.92 and the P-value of F-test is 0.00. Since P-value is smaller than 5% level of significance, we can reject the null hypo-thesis that all the partial slope coefficients are simul-taneously equal to zero. So the above estimated model is statistically significant.

The main objective of this study is to assess the socio-economic factors that affect undergraduate students academic performance in Bangladesh. The findings of the study show that students gender, class attendance, study hour, previous academic results, university ad-mission test score, family income, fathers education, mothers education, participation in part time job, students internet using for education purpose, parti-cipation in extra-curriculum activities and participation in group study are important determinants of students academic performance or CGPA (Cumulative grade point average). Among them, class attendance, study hour, previous academic results, university admission test score, family income, fathers education, mothers education, internet using for education purpose and participation in group study have  positive impact on CGPA (Cumulative grade point average). Dummy variable gender with reference category of being female students, participation in part time job and participation in extra-curriculum activities have nega-tive impact on CGPA. Although gender, participation in part time-job, internet using for education purpose and participation in extra-curriculum activities each have an impact on CGPA (Cumulative grade point average) , these impacts are not statistically significant in this study. Depending on the results of this paper, some of the following recommendations are given below in order to improve academic achievement of undergraduate students in Bangladesh. Since family income is one of the important determinants of stu-dents academic performance, policy should be taken to provide special support to those students coming from low income family. Working in the ground with school and college level and ensuring better edu-cational basement of students can play a significant role in improving academic performance of graduation level. Students should be careful to their regular class and study. Teachers can motivate their students in this regard by explaining the importance of regular class attendance and per day study hour. Since parents education has an important impact on improving academic performance of their children, adult edu-cation can play a significant role in this regard. Stud-ents should participate in their group study so that they can solve their academic query by discussing each other. Group study helps to understand academic lessons clearly which is important to improve gra-duation level academic performance. Moreover, the provision of teaching and learning materials, impro-vement of academic infrastructures, better teacher training and payment of incentive to teachers, ICT based education system and free internet access for both teachers and students are recommended to im-prove the quality of education. Understanding the social and economic determinants of undergraduate students academic performance is of great importance in policy formulation and educational planning poli-cies. Because incorrect assumptions leads to formul-ation of unsound policies on fees, scholarship and subsidies. Therefore, the outcome of this study can be useful to formulating policies and guidelines for over-all attainment of the education status of this country.

The author expresses his gratitude to the respondents for their co-operation in providing the required data for this study.

The author declares no possible conflict of interest in concerning the study, data collection and analysis.