Information and Students´ Preferences for Higher Education: Evidence from a Randomized Experiment in Colombia

We run the experiment on low-income Colombian students in 10th grade. The intervention consists of 9 statements presenting factual information regarding higher education (HE). The first 7 statements are the same for both treatment groups. However, the last 2 statement differ for each group: treatment 1 emphasizes on HE returns and enrollment rates in accredited universities, whereas treatment 2 emphasizes on financial aid and university costs in accredited universities.

2018-07-31

2019-02-15

Shift on preference for Higher Education (HE) attributes due to information treatments.
Importance of attibutes in HE decisions in the presence or absent of information.
Willingness to pay (WTP) to obtain better attribute levels in the presence or absent of information.

We run the experiment on low-income Colombian students in 10th grade. We chose a stratified random sample of Colombian schools which was representative of the population of schools across a number of dimensions (i.e. % of eligible for need-base scholarships, proportion of female, full-day at school and conflict victim.). We randomized the schools in 3 different groups, checking balance for the above-mentioned characteristics. We left out from the sample those schools for which contact information was not available. We have reached out 99 schools with at least one valid contact information (land phone, cellphone or email). The intervention consists of 9 statements presenting factual information regarding higher education (HE). The first 7 statements are the same for both treatment groups. However, the last 2 statement differ for each group: treatment 1 emphasizes on HE returns and enrollment rates in accredited universities, whereas treatment 2 emphasizes on financial aid and university costs in accredited universities. To built the statements, we gathered information from official government websites, focusing on statistics which are representative of the low-income student population. Likewise, to built the DCE, we: analized previous surveys of senior high school students, collected information from non-structural interview to 2 professionals working at the Ministry of Education, performed two focus groups in low-income schools with students in 10th grade, and tested final survey with freshman students at college.

We run the experiment on low-income Colombian students in 10th grade. We chose a stratified random sample of Colombian schools which was representative of the population of schools across a number of dimensions (i.e. % of eligible for need-base scholarships, proportion of female, full-day at school and conflict victim.). We randomized the schools in 3 different groups, checking balance for the above-mentioned characteristics. We left out from the sample those schools for which contact information was not available. We have reached out 99 schools with at least one valid contact information (land phone, cellphone or email). The intervention consists of 9 statements presenting factual information regarding higher education (HE). The first 7 statements are the same for both treatment groups. However, the last 2 statement differ for each group: treatment 1 emphasizes on HE returns and enrollment rates in accredited universities, whereas treatment 2 emphasizes on financial aid and university costs in accredited universities. To built the statements, we gathered information from official government websites, focusing on statistics which are representative of the low-income student population. Likewise, to built the DCE, we: analized previous surveys of senior high school students, collected information from non-structural interview to 2 professionals working at the Ministry of Education, performed two focus groups in low-income schools with students in 10th grade, and tested final survey with freshman students at college.

Thinking in increasing the chances that students actually “absorb” the statements, we adopted a “myth or reality” format, so far not presented in previous literature. Students were given a statement and then have to vote “myth” or “reality” according to their beliefs. Right after voting each item, they were told if the answer was right o not and showed additional information and/or explanation of the statement (e.g. statement: myth or reality? Less than 25% of the credits given by the ICETEX –governmental entity in charge of credits for HE- are given to low-income students. Explanation: Myth! 69% of students’ beneficiated from ICETEX loans are low-income students).

The information presented to the students were gathered from official government websites. We were careful in presenting statements according to their socioeconomic status (i.e. credits rules and returns differ according to socioeconomic characteristics). Statements tended to be general information as opposed to particular information (i.e. we say: in general, accredited universities tend to have higher quality than non-accredited; as supposed to X university tend to be better quality than Y).

Randomization was done by a computer, checking balance on schools observable characteristics.

Schools

Yes

99 schools (and 205 classes)

2553 students We performed a power analysis in order to plan the minimum number of observations using “optimal design software”. With a planned number of 100 clusters, power parameter equal 0.9 and type I error of 0.05, it is required on average 5 students per cluster. This also implies a minimum sample size of 500 students.

Control group =34 schools
Treatment 1 group=35 schools
Treatment 2 group=30 schools