Socioeconomic conditions and academic performance in higher education in Colombia during the pandemic

Abstract This article studies the relationship between the socioeconomic conditions of higher education students in Colombia and their academic performance during the pandemic. The household’s socioeconomic conditions are approximated by the education level of the parents, their occupation and the possession of assets. A multiple regression model tests the effect of these variables on academic performance before and during the pandemic. Results suggest that before the pandemic, the mother’s graduate education and household technology assets showed a positive impact on test score. Mixed effects of parents’ occupations by gender were also found. During the pandemic, the effect of the mother’s education remained the same, and the effect of technological assets, in-person education and high-quality accredited establishment increased.


Introduction
The pandemic has affected many dimensions of people's lives having obvious economic, employment and educational effects. The loss of employment and the reduction in income, together with greater job instability and uncertainty, have led households to face an extreme situation that often deteriorates their economic conditions. In the education sector, the pandemic has transformed the pedagogical, didactic and evaluative components of teaching and learning processes at all levels. Using virtual environments for learning, accessing information and as communication technologies have become increasingly important.
Although the socioeconomic conditions of the home may in principle have a greater influence on the academic performance of primary and secondary students, they are also relevant in higher education. Students with low economic conditions can be affected not only by their contextual factors but also by material conditions that can lead to dropouts, poor educational and work activities and low academic performance. The COVID-19 pandemic modified economic and social environments and can have important implications for their performance. Recent systematic reviews of literature and international empirical evidence suggest a positive relationship between a student's socioeconomic level and academic performance in higher education (Sirin, 2005;Schneider & Preckel, 2017;Rodr ıguez-Hern andez et al., 2020).
Studies analysing the effects of COVID-19 on higher education have come to a consensus that the coronavirus pandemic has contributed to strengthening the importance of contextual differences, ranging from connectivity problems to the amount and quality of technological tools available to students, teachers and higher education institutions to guarantee the continuity of the teaching and learning processes (C aceres et al., 2020;Cao et al., 2020). Consequently, although both teachers and students have identified the need to develop self-learning, autonomy and socio-emotional skills as an important challenge (Miguel, 2020), the most significant challenge was to continue providing both adequate and quality education seeking strategies that allow it to solve the problems posed by the digital divide and the socioeconomic conditions of students and their families.
Works, such as those of Alharbi (2020) and Bedoya et al. (2021), who have addressed the financial support of universities and states to students in times of pandemic, emphasise the economic support that universities have provided to their students. Public universities have provided aid in the form of financing, tuition discounts and subsidy increases. Private universities have offered tuition discounts and extensions of payment dates (Bedoya et al., 2021). Although these grants have been useful, university dropout rates are expected to increase because COVID-19 has resulted in both students and parents losing the jobs that guarantee their support and those of other family members. In addition to the health crisis, education, including universities, was one of the most affected sectors (Guti errez et al., 2019;C aceres et al., 2020;Tejedor et al., 2020). Technology-mediated education, which everyone had to take on as the only alternative to keep the university running, not only created rapid and challenging learning efforts for students and teachers but also made unequal and disadvantageous learning spaces more visible for students from poor families (Cervantes & Guti errez, 2020;Rodr ıguez, 2020;Tejedor et al., 2020).
From a social point of view, the impact of COVID-19 in Colombia has been enormous. Recent studies suggest that the impact of COVID-19 would lead to losing much of the progress achieved in reducing poverty and inequality over the last few decades. According to C ardenas and Zuleta (2020) and N uñez (2020), the most affected people will be those in the middle and lower middle-income households whose incomes are more dependent on jobs associated with the sectors most vulnerable to the current crisis. So, the effect of the pandemic is especially strong in economies such as Colombia's, where there is a strong relationship between income levels and vulnerability of income sources to the current situation.
Regarding tertiary education, Mora (2016) emphasised that higher education operates as a producer and reproductive field of exclusion and inequality in Colombia. Self-selection processes operate as mechanisms of self-exclusion of those who meet all the formal requirements to access higher education but who do not attempt to do so because they consider that the quality of the education they have received or their socioeconomic conditions will prevent them from successfully completing their educational process (Mora, 2016;Corredor et al., 2020). Gallardo et al. (2011) concluded that the higher education system in Colombia converts social privileges into individual merits, thus justifying and reproducing inequalities of class, ethnicity, gender and regional location. Cuenca (2016) argued that the fact that quality educational opportunities are not available to all individuals, regardless of their position of origin, deserves a reflection about the role of public policy. This issue is relevant because during the social protest carried out in Colombia in 2019, young people demanded better job opportunities and enhanced quality higher education from the state (Acevedo & Correa, 2021;Garc ıa & Arias, 2021). According to Castillo and Garc ıa (2019), in Colombia young people between 18 and 30 years old face special obstacles in getting a formal job because labour indicators, such as unemployment, job insecurity or income, are significantly worse in this group than in people over 30 years old.
Consequently, the aim of this article is to evaluate how the COVID-19 pandemic affected the incidence of students' socioeconomic conditions on academic performance in higher education in Colombia. To accomplish this, a multiple regression approach estimates the impact of socioeconomic variables such as parental education and occupation and the existence of assets in the home on the score in the Instituto Colombiano para la Evaluaci on de la Educaci on (ICFES) Saber Pro tests before and during the pandemic period. The model includes students' sociodemographic characteristics, conditions related to their home and characteristics of the higher education institution where they study.
According to the literature, higher education has shown a positive but weak relationship between academic performance and the socioeconomic situation of students (Rodr ıguez-Hern andez et al., 2020). However, these studies remain useful especially in countries such as Colombia with many inequalities and large social gaps between the rich and poor (Carreño et al., 2017;Murillo & Graña, 2020;Palacios, 2020). These studies have focused on analysing the effects of families' economic situations on the academic performance of students (Murillo et al., 2016;Zhou et al., 2016;Gorard & Siddiqui, 2018;Kr€ uger, 2019; Organisation for Economic Co-operation and Development (OECD), 2019; Janmaat, 2022). In some cases, they have sought to influence the design and implementation of public policies in education.
To quantify and understand the socioeconomic conditions of student households, literature has considered home ownership, loans to acquire, materials, extension, structure, access to public services, number of inhabitants, furniture, electrical appliances, vehicles and materials for school use (Cowan et al., 2012;De Clercq et al., 2017;Charles et al., 2018;Johnson, 2020). Having these possessions at home provides valuable information about family income over short and long periods. They are considered relevant to determining the effect of income on the academic performance of students (Shavit et al., 2007;Charles et al., 2018;Johnson, 2020). Additional variables such as the education levels of the parents and their occupations are also important because they also directly relate to family income and the support and accompaniment that students can receive from their parents on school tasks (Crawford, 2014;Crawford et al., 2016;Erola et al., 2016;Smith, 2016;Jacobs & Wolbers, 2018). These variables are contextual and explanatory to help configure more complex interpretations of the relationships between the economic situation and academic performance.
Studies with the most theoretical and empirical robustness in recent times performed meta-analyses in these two categories. These studies emphasised overcoming the approach of using socioeconomic situation as a covariate, instead opting for more comprehensive readings to determine what influence the association between academic performance and socioeconomic situation had on student experiences and results (McKenzie & Schweitzer, 2001;Walpole, 2003). Similarly, Rodr ıguez-Hern andez et al. (2020), Westrick et al. (2015), Richardson et al. (2012) and Schneider & Preckel (2017) conducted meta-analyses to obtain a deeper understanding of the relationship between socioeconomic status and academic performance. These metaanalyses were necessary to adequately address the growing diversity in the student population and formulate public policies that close the gaps between rich and poor in countries characterised by substantial economic and social inequity.
Colombian studies have considered the following as socioeconomic variables: income levels, mother's education, unemployment rates in the region, number and position among siblings, percentage of students that are head of household (who have the responsibility of managing the household and making decisions), percentage of students whose mothers have professional and postgraduate degrees, as well as percentage of students who have financed their studies with a scholarship (Melo et al., 2017). Studies show that socioeconomic variables affect not only cognitive variables such as access to information and proper development environments but also noncognitive variables, since more educated parents and higher income can stimulate children to develop extracurricular activities that promote school learning (Escobar & Orduz, 2013;Brito & Palacio, 2016;Melo et al., 2017).
Accordingly, the literature suggests that economic factors are a key determinant of students' academic performance in higher education. The most recent studies carried out during the pandemic show that the differences in economic conditions, such as having access to technological tools, have generated an increased gap in student performance that has led governments to the same educational institutions to implement strategies to offset these negative effects. Under these conditions, providing evidence of this relationship in developing countries that tend to present large gaps in the economic and educational spheres enriches the development of this literature.

Methodology
A multiple regression econometric model delineates the relationship between students' academic performance and their socioeconomic conditions. The dependent variable of academic performance in higher education is explained by a set of independent variables related to the socioeconomic conditions of the home, the sociodemographic characteristics of the student and the characteristics of the academic programme and the educational institution. Equation (1) presents the linear econometric model.
This equation approximates academic performance by the Score obtained in the Saber Pro test. The explanatory variables of the model are grouped and presented in matrices to explicitly distinguish the coefficients of interest to evaluate the interaction effects between explanatory variables. In this case, a refers to the intercept of the equation (i.e., the mean value of the score when all explanatory variables are equal to zero) and b, c, s, h, d, p, r, q and u are vectors of coefficients that refer to marginal effects of explanatory variables on the score (that is, the marginal change in the score when one explanatory variable varies in one unit).
The explanatory variables are matrices containing the socioeconomic conditions of the household. The W matrix contains columns of the education level of both mother and father (primary, secondary, university). The X matrix contains the occupation of both mother and father (enterprise owner, professional, pensioned, self-employed, administrative assistant, machine operator, sales, industrial or domestic cleaner and housewives, among others. The Y matrix contains household assets, such as the existence of a television, a washing machine and a car, the possession of a computer and the existence of internet access. These variables are all qualitative and equal 1 when the household has the asset and zero otherwise. Education levels of the parents and their occupation are included based on the works of Smith (2016) (2020) and Charles et al. (2018). Computer ownership and internet access are considered based on the works of C aceres et al. (2020) and Cao et al. (2020).
The Z matrix contains columns of variables related to demographic and socioeconomic characteristics of the student, including age, sex, number of hours worked, level of overcrowding in the home and socioeconomic strata, together with the characteristics of the programme as a study modality and of the institution, such as nature, type of institution and institutional accreditation. Socioeconomic strata are a categorical variable calculated by government as a proxy of the level of the household's wealth in the urban sector. This variable ranges from 1 to 6 with households in strata 1 having, on average, less wealth than households in strata 6. Rural households do not have socioeconomic strata. Including these variables is based on the literature previously presented and the availability of information in the database considered. Table 1 lists the variables considered in the econometric model.
Since the goal is to evaluate the changes in the impact of socioeconomic variables during the pandemic period, Equation (1) includes a dichotomous variable t that distinguishes the observations in two time periods: before and during the pandemic. The variable t has the value 0 for the year 2019 and 1 for the year 2020. This variable generates interaction effects in the model, multiplying it with the values of all explanatory variables and generating new matrices denoted with an asterisk. For example, W Ã is a matrix that contains all values of variables only in 2020 (when t ¼ 1).
Coefficients p, r, q and u are the multiplicative effects between the sets of variables and the variable t. These coefficients give evidence of the direction of change in the incidence of the explanatory variables considered during the pandemic period. The statistical significance of these coefficients implies that these factors changed during the pandemic period. Lastly, the variable l is the disturbance term of the model (i.e., a random effect not explained by the model).
The coefficients of the model are estimated by the Ordinary Least Squares-OLS method. The model contains both qualitative and quantitative explanatory variables. Categories of qualitative variables are incorporating as dummy variables. Coefficients of these categories are differential effects of the base category. For example, the coefficient of 'having a mother with postgraduate education' reflects the differences in score of this feature when 'having a mother with no education' is omitted. The estimates do not contain bias corrections as a result of endogeneity problems in the explanatory variables. Therefore, the estimates are not causal effects.

Data
The data used in this work comes from the database of the standardised ICFES Saber Pro state tests taken by students in the last semesters of their tertiary education in Colombia during the years 2019 and 2020. This database has the sociodemographic and socioeconomic status of the students and their homes and provides information about the nature of the institution and the modality of the academic programme of study. It contains more than 200,000 observations per year.

Descriptive statistics
In 2019, the average global score stood at 148.02, while in 2020, both its average value and its deviation increased. The average student age was 24.7 years and, as expected, it did not show a major change (Table 2). Most of the students who took the test were women (Table 3). Most of the students worked and more than a third were employed for more than 30 hours per week. More than half the students shared a bathroom with more than three people and the vast majority were from socioeconomic strata 1, 2 and 3. During the pandemic, there was greater participation from the lower strata. More than 80% of the students had a TV, computer, internet and washing machine at home while only 36% had a car. For 2020, the increase in computer and internet ownership stands out.
There were a greater number of mothers completing secondary, technological, higher and postgraduate educational cycles in both periods (Table 4). During the pandemic, although both groups increased their participation at professional and postgraduate educational levels, women showed the greatest increase.  Most of the higher education took place as in-person programmes at universities, with 42% having high quality accreditation. By 2020, there were small increases in virtual distance programmes and private institutions (Table 5).

Econometric results
Econometric results are presented in four sub-sections according to the aim of the paper and the econometric model in Equation 1. First, coefficients b and p show the incidence of parental education on academic performance before and during the pandemic respectively. Second, coefficients c and r show the incidence of parental education in the same periods. Third, coefficients s and q show the incidence of household assets before and during the pandemic. Finally, coefficients h, u and d relate the incidence of control variables associated with socio-demographic and socioeconomic features of the student before and during the pandemic, as well as with the dummy of year.

Parental education
Econometric results exhibit the estimated coefficients of the effect of the level of education of both mother and father on academic performance  before and during the pandemic (Table 6). In this case, results are only for levels of education greater than secondary. All the results obtained for the pandemic period are noted by the term 'Time multiplied by' the respective category. Coefficients are accompanied by its statistical significance specified by p-value and by its standard errors that appear below each coefficient in parenthesis.
According to Table 6, in 2019, mothers' education had a greater impact on the academic performance of the students than the fathers' education. While having a mother with a graduate degree represents an academic performance of 0.46 standard deviations higher than having a mother without any education, in the case of the father, this coefficient is only 0.25. This pattern, where the effect of the mother's education is higher than the effect of the education of the father, is maintained for all education levels considered and is in line with that reported in the international literature (Ru ız de Miguel, 2001;Erola et al., 2016).
In 2020, the coefficients for university studies, university degree and postgraduate degree did not vary for mothers but fell for fathers with postgraduate degrees and those who had completed higher education. This implies that during the pandemic period, mothers' education played a very important role in student test scores since, unlike other effects considered in this study, they stayed at their initial levels (Comisi on Econ omica para Am erica Latina y el Caribe (CEPAL), 2020; Comisi on Econ omica para Am erica Latina y el Caribe (CEPAL) & Organizaci on de las Naciones Unidas para la Educaci on la Ciencia y la Cultura (UNESCO), 2020).

Parental occupation
Results suggest differentiated effects of parent occupations according to gender of the parent (Table 7). In 2019, all mothers' occupations except big enterprise owners, report positive associations with academic performance. In the cleaner, machine operators and home occupations, the coefficients report impacts of 0.17, 0.15 and 0.12 standard deviations respectively This suggest that students with mothers in such occupations are related with higher test scores compared to students with mothers in agricultural activities, which is the base category. In contrary, most of the fathers' occupations report a negative coefficient except for self-employed and operator. Most of these effects are marginal.
The differences of the occupational effects by gender could be associated with the fact that mothers in the labour market may have unobserved noncognitive abilities that have a positive effect on a student's performance. Abilities such as motivation, perseverance and discipline can shape the relationships within the households and influence students' behaviour. Mothers with the 'home' occupation also report a positive impact on students' performance, suggesting their key role in providing good conditions and supporting the learning process. During the pandemic, the effect of the mother's occupations remains. Students with mothers in operator, selfemployed and home occupations exhibit increases in their effect. (The interaction coefficient between variable and dummy time is positive). This result highlights the value of mothers both at home and in the labour market even under extreme efforts on academic performance.

Household assets
Having a television and a motorcycle is associated with lower academic performance, while having a computer and the internet positively affects the score (Table 8). The effect of the latter two on academic performance is largely quantitative in nature. During the pandemic period, as expected, the correlation of having a computer and internet access with academic performance increased.

Demographic, household and institutional characteristics
In 2019, being a woman and being older were associated with lower test scores (Table 9). From 2019 to 2020, this gap widened in both cases, meaning the pandemic increased the inequality between these socio-demographic groups. In 2019, participants in the highest socioeconomic strata had higher test scores and higher levels of overcrowding in the household reduced test scores. In 2020, the stratum variable did not show any changes but the gap increased for people who shared a bathroom with 3 or 4 others.
Higher test scores are associated with belonging to an in-person education programme, being from a university, being in a public institution and having a high-quality accreditation. In-person education, university and highquality accreditation coefficients increased by 0.02 standard deviations in 2020. This increase in the gaps between study modalities and the characteristics of the institutions reflects the importance of these quality factors during the pandemic. According to the coefficient of time dummy, average academic performance increased by 0.39 standard deviations during the pandemic, which means that the average score in 2020 was higher than in 2019.

Socioeconomic conditions and academic performance in the pandemic
The results of this article suggest that the high-level of education of the mothers, occupations of self-employed and operators for both parents and cleaner and being at home for the mothers, together with the possession of a computer and access to the internet, had a positive influence on the academic performance of the students in 2019. Likewise, belonging to an in-person programme modality, studying at a public institution and being both in a university and a high-quality accredited establishment are associated with higher test scores. During the pandemic, the effect of a mother's education and occupations remains. Students with mothers in operator occupations, self-employed and those at home, exhibit increases in their effect. Technological assets, as expected, are important, the correlation of having a computer and internet access with academic performance increased by 2020. In the same way, inperson education and high-quality accredited establishments are associated with higher scores during the pandemic.
Results confirm the validity and importance of the studies that consider the relationship between the economic condition of families and the academic performance of students (Zambrano Vera et al., 2019;Rodr ıguez-Hern andez et al., 2020). In the specific case of higher education and in line with Tejedor et al.,  Cao et al. (2020), the results show that the economic effects of the pandemic will deepen existing educational inequalities, which in turn can also translate into a slowing of reaching goals projected for higher education.

2020) and
Although the COVID-19 pandemic has had negative effects on higher education in Colombia, these effects may be an opportunity to strengthen the pedagogical and technological aspects of higher education educational institutions (Miguel, 2020;Ordorika, 2020). In line with Tejedor et al., (2020) these actions should strengthen public policy to promote the permanence of students from families with low incomes and, therefore, decrease the drop-out rate. Actions can also be aimed at strengthening virtual and hybrid modalities in universities and could be accompanied by a necessary reflection on the low quality of distance programmes compared to in-person programmes (Bedoya et al., 2021). This implies pedagogical reflections and concrete actions on the conditions that have generated a gap between in-person and virtual education in the face of the future of higher education (Ram ırez, 2016;Cobo, 2017).
One of the outstanding mechanisms to reduce economic inequality in the medium and long term is public education. The opportunity to access and obtain a professional degree in a tertiary public institution improves future incomes and contributes to reducing socioeconomic gaps between groups. In Colombia, the financial strengthening of such institutions to enhance factors such as professors, infrastructure (physical and technological) and research directly affect their quality. As results suggest in Table 9, before the pandemic, students from public institutions obtained 0.11 standard deviations more than those in the private sectors. The fact that during the pandemic this coefficient did not increase shows how difficult it was for such institutions to adapt to the new context. This is why the sector needs attention.
In line with Montes and Lerner (2012) and Karagiannaki (2017), a household's physical and material conditions also make a difference in students' academic performance. The need for materials such as a computer or internet access is part of the wealth mentioned (Sirin, 2005;Lovenheim, 2011;Cowan et al., 2012;De Clercq et al., 2017). Not having these materials makes it more difficult to achieve academic success and puts students who come from poor families at a great disadvantage. Computers and internet can no longer be considered sumptuous but essential for learning. Therefore, higher education institutions must continue to make great efforts to reduce any digital and technological gaps between students.
Additionally, since the percentage of students who must work to pay for their studies (Melguizo et al., 2016;Palacios 2020) and the youth unemployment figures during the pandemic (Departamento Administrativo Nacional de Estad ısticas (DANE) 2021) increased, it is also necessary to pay attention to the employment situation of young university students. The loss of work can mean drop out for those students whose education and even part of their personal and family support depend on the student's work income (De La Hoz et al., 2012). Literature also suggests that students having jobs that are more than part-time and associated with their career not only affects their academic performance positively but the additional experience also enables better job placement after graduation (Bezerra et al., 2009;Yanbarisova, 2015;Palacios, 2020). Therefore, two complementary strategies should be considered. Government programmes such as 'Young State' should be strengthened to include more students. This type of programme provides students the opportunity to acquire experience in the public sector and facilitates engagement in the labour market after graduation. At the university level, improving institutional efforts through 'employment offices' that bring students closer to the productive public sector is key to reducing the economic and time costs of finding a job.

Conclusions
This work studied the relationship between the socioeconomic conditions of students and their academic performance before and during the pandemic caused by COVID-19. A multiple regression model evaluated the effects before and during the pandemic. Results show before the pandemic, a positive impact of the mother's graduate education, mixed effects of parents' occupations and higher scores for students who have computer and internet at home, belong to an in-person programme modality, study at a public institution and are both at a university and a high-quality accredited establishment. During the pandemic, the effect of technological assets, in-person education and belonging to a university and high quality accredited establishment increased. The positive effect of the mother's education remained by 2020.
The results highlight how important socioeconomic variables are in determining academic performance in a pandemic. The economic gaps generated by the pandemic translate into technological and educational gaps that in turn will produce greater economic and social gaps in the future. In this sense, it is necessary to design policies that compensate for the adverse effects of widening economic gaps on educational gaps. These policies need multidimensional approaches that not only consider the strictly economic 'at-the-family' level but also proper conditions from the financial, technological and pedagogical points of view on the part of higher education institutions.
Policy definition will require more detailed studies at the territorial level of the effects of the pandemic on the economic conditions of households and its impact on academic performance. The pandemic has had heterogeneous effects not only demographically but also spatially. Deepening the understanding of such gaps and their impact on the educational and employment trajectories of specific populations makes up an important line of research to strengthen in this field.

Disclosure statement
No potential conflict of interest was reported by the author(s).