Private Tutoring and Academic Achievement in a Selective Education System

AN0186774485;ede01aug.25;2025Jul23.02:57;v2.2.500

Private tutoring and academic achievement in a selective education system 

This study explores how private tutoring relates to students' transitions to demanding post-compulsory schools and their success there, considering their competencies after tutoring but before the transition. Analyzing PISA and linked register data from Switzerland, we find that students who received private tutoring before the transition are more likely to struggle in selective schools compared to non-tutored peers with similar competencies. While our results are not causal due to the non-random nature of private tutoring uptake, our findings underscore a potential concern regarding selection mechanisms for entry into selective education.

Keywords: Private tutoring; educational achievement; PISA; Switzerland

1. Introduction

In most countries, access to prestigious, high-quality schools at upper-secondary level is selective and restrictive. In order to positively influence the chances of their children to access such schools, parents often organize private tutoring for their children.[1] Therefore, it is not surprising that while private tutoring is widely prevalent in many countries all over the world (Hille, Spieß, and Staneva [14]; Park et al. [20]), participation rates are higher in countries with a highly selective school system (Bray [5]; Guill and Lintorf [11]). While there is a growing, although not abundant, literature about the causal impact of private tutoring on competencies (Behrman et al. [4]; Chih-Hao [6]; Cole [7]; Fryer and Howard-Noveck [10]; Guill, Lu¨dtke, and K¨oller [12]; Guo et al. [13]; Hof [15]; Zhang [22]; Zhang and Xie [23]), there is – to our knowledge – still a lack of knowledge about the mid- and long-term effects of private tutoring.

In this paper, we fill some of this gap by looking at the impact of private tutoring on successfully finishing a selective academic school at the upper-secondary level (baccalaureate schools). We are not addressing the question of whether the use of private tutoring increases the competencies of the students[2] but rather look at the success rates of students of comparable pre-transition competencies, comparing those who entered the schools without private tutoring lessons to those who had made use of private tutoring before entering. Hence, if we assume that private tutoring has a positive impact on competencies, we compare the success probability of students who had reached a certain competence level without private tutoring to students who had reached the same levels only thanks to private tutoring. If, however, the impact of private tutoring is modest or negligible, we just compare students with similar abilities with and without private tutoring.

From the point of view of educational policy-making, this paper is of interest because the admission processes to selective schools are all explicitly or implicitly designed in a way to predict a students' academic potential, which should be linked to the probability of successfully passing the baccalaureate school. Independent of whether grade point averages (GPA) of previous school years, teacher recommendations or standardized external tests are used to make the entry selection, and these systems all have in common that those students who pass the entry threshold are seen fit to pass the selective upper-secondary academic schools. However, some students may have passed the thresholds only by taking extra training (private tutoring), and others have passed the thresholds without any extra support. This difference – and the information contained therein – is not taken into account when the educational authorities make their selection, but it could be predictive for future academic success. Our paper addresses the question of whether students of comparable competencies with and without prior private tutoring really have the same success probabilities in a selective academic program.

To address this research question, we used PISA (Program for International Student Assessment) test data from 2012 from over 12'000 Swiss students, merged with register data. This allows us to follow the educational achievements of these students after the PISA test was taken. We also make use of the fact that in 2012, as a national option, Switzerland included a detailed survey on private tutoring in the PISA questionnaire. PISA is a low-stake exam as the outcome does not matter for the student taking it. Learning to the test will, therefore, not be an issue. As we focus on the student's skills – a snapshot of where the student is – a low-stakes exam is preferable to a high-stake one, as the students will not make any extra effort to raise their scores (Akyol, Krishna, and Wang [1]). However, we have to assume that students with and without private tutoring do not differ in their attitude to taking a low-stake test and that individual test scores are not biased in a systematic way.

Our empirical results indicate that it matters for the success rate in a selective program how a certain competence level before entering the program was reached. Students with private tutoring do worse in baccalaureate schools, i.e. they have a lower probability of receiving a diploma in the baccalaureate school, within the regular time frame and overall. The results highlight the fact that passing a particular threshold at the time of selection does not reveal all the necessary information to predict the future academic ability of students.

This paper is structured as follows. Section 2 describes the Swiss education system. Section 3 describes the data, while Section 4 introduces the empirical strategy. Section 5 reports the results, and Section 6 presents the conclusion.

2. Swiss education system

Compulsory education in Switzerland comprises kindergarten plus nine years of schooling, with six years of primary school and three years of lower secondary school. At the lower secondary school level, different school-type models exist that vary from region (canton) to region. The majority of school-type models sort students into different school tracks according to their intellectual abilities. Although two to four different tracks exist, the majority of regions apply a three-track model: an upper-level school track (pre-baccalaureate), which teaches the more intellectually demanding courses, an advanced level school track, and finally, one offering basic-level courses.

After finishing compulsory schooling (9th grade), students can choose between two different possibilities: full-time general schooling (baccalaureate school preparing for university) or vocational schooling, of which most of the programs are so-called dual (school plus firm-based training; apprenticeships). Approximately 20% of school graduates attend an academic baccalaureate school, which prepares them for university. All baccalaureate schools in Switzerland lead to nation-wide free access to all universities. However, there exists neither a national leaving exam at the end of compulsory education nor a national entrance exam for baccalaureate schools.

The Swiss education system comprises different regions with independent educational policies, which leads to a considerable variation in the education options within compulsory education (tracking) as well as in how baccalaureate schools and access to them are organized. Requirements and entry procedures for admission to baccalaureate schools differ considerably by the track in compulsory education and regional rules, but in simplified terms, we can identify three types of entry procedures. Type 1 concerns students who are pre-tracked for baccalaureate schools and requires the students mainly to reach a sufficient GPA in order to transition into the post-compulsory baccalaureate school. Type 2 (for not pre-tracked students) characteristically does not entail a leaving or entrance exam but relies instead on previous attainment records (Grade Point Average, GPA) and/or the teachers' evaluation of the students. Under Type 3, a leaving or entrance exam usually determines admission to the baccalaureate school, comparable to a standardized achievement test (SAT). Several regions offer Type 1 combined with either Type 2 or 3.

3. Data

We match data from the PISA 2012 survey to register data on student enrollment in the Swiss educational system.[3] This allows us to follow the PISA participants' transition into upper-secondary education and their success within their chosen educational path. The register data is provided by the Federal Statistical Office Switzerland and contains yearly information on all students who are enrolled in any type of educational institution. We follow students for seven years after their compulsory education. We register whether a student receives a diploma from a baccalaureate school and whether this degree is reached in the nominal study duration or with a delay.

We use two different samples for the different stages of our analysis: In the first step, we analyze whether the students enter a baccalaureate school, using the transition sample, which includes all students in the PISA study. In the second step, we investigate whether students in baccalaureate schools finish school successfully within the seven years we observe. For this analysis, we focus only on students, who did transition into a baccalaureate school; the baccalaureate sample.

Due to regional oversampling of students in the last year of compulsory school in the PISA study, we observe a large sample of 12'696 graduates from compulsory school in 2012 and a sample of 3055 students who started studying at a baccalaureate school. In addition to the international student background questionnaire, the Swiss PISA 2012 questionnaire included a detailed survey on private tutoring as a national option.[4] The survey focused on self-initiated tutoring, i.e. private tutoring not initiated by the school or some official party but rather by the student's parents or the student him or herself. Furthermore, it is fee-based private tutoring in academic subjects.[5] The questions provide information about the frequencies, motives and other relevant variables related to private tutoring in 8/9th grade among 9th graders.[6] The combination of this additional information, the PISA scores and the register data allows us to analyze whether students who had obtained the same PISA scores and are comparable in many other ways (the PISA background survey provides us with a rich set of observable student characteristics) but differ in relation of making use of private tutoring prior to the PISA test, have similar or different success receiving an academic baccalaureate school, conditional on transitioning into this educational track. We code the variable private tutoring as a binary variable if the student had any private tutoring in the 8/9th grade. As a robustness check, we additionally create a variable that captures only private tutoring that took place on a regular basis, e.g. tutoring over several weeks or months (as a distinction to private tutoring on an irregular basis, e.g. tutoring during some lessons, which is included in the first binary variable).

The PISA data set further provides data on the achievement scores in mathematics, language and science, which we use as proxies for cognitive skills. The focus of the 2012 PISA test was on mathematics; therefore, the math scores are the most precisely measured test scores. Furthermore, not all students have had a language or science test, which is why we focus on math skills in our main analysis. It is important to note again that these skills are measured after the tutoring has taken place. We also use the OECD thresholds to categorize students into competence levels from 1 (low) to 6 (high), which help to better understand the actual meaning of the continuous PISA scores (OECD [19]). Since we investigate the transition into and success in baccalaureate schools, which are academically demanding, we are particularly interested in high-performing students (PISA competence level 4–6). While the PISA test as such is not used for the transition decision, the competencies that are demanded for the transition into a baccalaureate school closely correspond to a PISA competence level of at least 4 in all subjects (SKBF [21], 150).

The PISA student survey elicits whether the students had been late for class during the weeks preceding the test. We use this information on punctuality as a proxy for non-cognitive skills[7], as a binary variable, with a value of 1 if never late for school.

In addition to the tutoring and skills variables, we include a number of socio-demographic and regional control variables. Most of the socio-demographic variables are taken from the PISA test. The migration back-ground indicates whether the student has been either born abroad or whose parents have not been born in Switzerland. We generate a dummy variable for missing migration background. Language is a binary variable that captures if the language the student speaks at home matches with the language of the PISA test.[8] Missing values are again captured by a dummy variable. The variable parents' education reports the highest education of any parent in the family in years of education. The socio-economic status is captured by the HISEI index for the parent with the higher-ranking occupation. For the regional controls, we use additional information from the statistical office in Switzerland. We include an indicator for the share of students in baccalaureate schools. Finally, we control for the language region with a dummy for the German-speaking part of Switzerland.

Table 1 reports the summary statistics for the most important variables for both samples, transition and baccalaureate. The samples are split into the groups of students who report to have had any paid private tutoring in the 8th or 9th grade and those who did not. We can observe that overall the students with tutoring are less likely to transition into a baccalaureate school, and fare worse on all cognitive and non-cognitive skills than the average of students who did not take private tutoring. This is not surprising because it can be assumed that students with lower competencies have a larger need for private tutoring and that they have on average worse competencies than other students even after private tutoring. This reverse causality, however, is not problematic for our study, since we conduct our analyses conditional on having the same competencies with or without private tutoring. We further observe that the parents' education is positively related to having tutoring. Living in the German-speaking part of Switzerland is negatively related to having had any tutoring. Among the students who do transition into a baccalaureate school, those who had tutoring have a lower rate of receiving a diploma. But also in this subsample, they have on average lower skills.

Table 1. Descriptive statistics.

Figure 1 illustrates the number of students with and without tutoring by math level. The baccalaureate sample is a more selective sample, in which the majority of students scored on the PISA math test above level 4.

Graph: Figure 1. Number of students with and without tutoring.

4. Empirical strategy

We start by estimating the marginal effects of tutoring on the probability of entering a baccalaureate school. The goal of this first step is to learn more about the relationship between tutoring and transition. However, we make no claims on the causality or even direction of the relation. In fact, it is likely that students who intend to enter a baccalaureate school are more inclined to seek tutoring, especially if their initial competencies are close to or below the minimally required skills level for the baccalaureate school (Guill and Lintorf [11]; Lorenz and Stubbe [18]). Nevertheless, documenting the transition process serves as an important initial stage in answering our main question on the later success.

For the students who did start a baccalaureate school, we then estimate the probability of finishing the program successfully and receive a diploma. Since the take-up of tutoring likely is not random and we cannot control for the factors that influence the take up decision, the results of our estimation cannot be interpreted as the causal impact of previous tutoring on success. However, we can learn more about the mismatch in the selection process, and the likely role tutoring plays in it. In addition to the average marginal effects of private tutoring, we are especially interested in the marginal effects of tutoring at different levels of math skills. Since the impact of tutoring, as well as the intentions behind it, likely differs between students along the skills distribution (Kang and Park [17]), we allow for an interaction between math skills and tutoring.[9] Using a probit regression, we estimate the following model:

Graph

P(y|x)=f(β0+β1Tutoring+β2Skills+β3Tutoring#Skills+γZ+ϵ),

where P (y) is either the probability of transitioning into a baccalaureate school or the probability of finishing a baccalaureate school with a diploma. Tutoring is a binary variable that captures any kind of paid private tutoring during grade 8 or 9.[10]Skills captures the competence level of a student, which we proxy with the PISA math score in the main specification. Z contains individual and regional characteristics. The residual is likely to be correlated for students in the same school, we, therefore, cluster on this level.

5. Results

5.1. Transition and success

Table 2 shows a negative average marginal effect of tutoring on the probability of transitioning into a baccalaureate school in column (1).

Table 2. Average marginal effects on transition and success probabilities.

1 Note: Marginal effects of probit estimates at sample means. Columns 1–2 provide the marginal effects on the probability to transition into a selective school for the complete sample. Columns 3–4 provide the marginal effects on the probability to succeed for the subsample of students who transitioned. The marginal effects of tutoring and math score in columns 2 and 4 for are based on the coefficient of the respective variable and an interaction term. We additionally include binary variables for missing values in language and migration background. Standard errors clustered at the school level. *** p