Perceptions and prospects in life sciences in a heterogeneous Latin American population

Particular challenges exist for science education in the developing world, where limited resources beget curricula designed to balance state-of-the-art knowledge with practical and political considerations in region-specific contexts. Project-based biology teaching is particularly difficult to execute due to high infrastructural costs and limited teacher training. Here, we report our results implementing short, challenging, and low-cost biology courses to high school and college students in Bolivia, designed and taught in collaboration between scientists from developed nations and local science instructors. We find our approach to be effective at transmitting advanced topics in disease modeling, microscopy, genome engineering, neuroscience, microbiology, and regenerative biology. Importantly, this approach was unaffected by the students’ backgrounds, education level, socioeconomic status, or initial interest in the course, and increased participants’ interest in pursuing scientific careers. These results demonstrate efficacy of participatory learning in a developing nation, and suggest that such techniques could drive scientific engagement in other developing economies.


INTRODUCTION 67
Latin America is home to more than 10% of the world's population, but 68 accounts for only 5% of the global scientific output [1][2][3][4], as measured by quantity 69 of original research publications and patent awards. A key indicator of 70 performance in science, technology, and innovation (STI), these output metrics 71 suggest significant underrepresentation of the region on the global stage. Even 72 outside of Latin America, Latinos tend to be underrepresented in scientific 73 careers. For example, in the United States, while Latinos comprise over 17% of 74 the country's population, they receive under 6% of graduate degrees in science, 75 technology, engineering, and mathematics (STEM) fields [5]. Importantly, while 76 Latinos are among the fastest growing demographics in the developed world, 77 their participation in higher education, and science in particular, is not projected 78 to keep pace with this growth [6]. Several intervention strategies have been 79 deployed in an attempt to minimize these gaps, but disproportionately low Latino 80 representation in STEM remains a conundrum for policymakers, universities, and 81 science educators. Prior to the start of the course, we held a team retreat in Bolivia. This 273 retreat was mandatory for all foreign and local instructors. In addition, we invited 274 experts to further train and discuss with instructors on project-based learning, as 275 well as challenges of teaching science in Bolivia. 276 277

Student recruitment 278
We used three different channels to recruit students: press, social media, 279 and visits to schools and universities. We opened applications on August 6th, 280 2017, Bolivian Independence Day, a major holiday and one of the top newspaper 281 circulation days. We coordinated a simultaneous press release in the top three 282 newspapers in the country: El Deber (Santa Cruz); Los Tiempos (Cochabamba) 283 and Página Siete (La Paz). Each newspaper wrote, free of charge, a two-page 284 article describing the program and inviting students to apply. Importantly, El 285 Deber newspaper made this article a cover article. In addition, we coordinated 286 appearances on several TV shows. Moreover, we orchestrated several other 287 articles, including interviews to instructors, in different press media throughout 288 the application period ending October 15th, 2017. 289 290 In social media, we created a Facebook page named "Clubes de Ciencia 291 Bolivia", which at the start of the application period had over 33,000 followers. 292 Page viewing statistics show that our online presence reached over 45% of  24-year-old Bolivians with Internet access every week, suggesting that this 294 approach was optimal to target a large audience. Working with local filmmakers, 295 we created videos for recruitment, featuring instructors introducing their course, 296 as well as multimedia presentations illustrating the potential for positive impact 297 from scientific engagement in national development (see Video S1). 298

299
In order to recruit additional students, particularly in rural areas, we 300 organized on-site visits to schools across the country. Our program had had two 301 previous editions and at the time had over 300 alumni, several of which volunteer 302 13 with the organization. These volunteers are based in 6 of 9 departments in 303 Bolivia. Volunteers were trained online and given a standard presentation. They 304 then visited schools and taught potential students how to apply. 305

306
In order to facilitate the recruitment of students, we incorporated an 307 application platform online, which could be accessed via cell phones or 308 computers. Cell phone penetrance in Bolivia is high. While no official numbers 309 are available, a recent study in hospitals of the cities of La Paz and El Alto found 310 that over 96% of patients ages 18-29 owned a cell phone [34]. Recent efforts 311 from the Bolivian government have delivered cell phone service in rural areas of 312 the country. Moreover, Quipus, a government-subsidized company has provided 313 computers to school all around Bolivia. Altogether, we believe that the 314 combination of press coverage, social media presence, and school visits 315 provided an easily accessible application platform ensured the targeting a 316 diverse student pool. 317 318

Selection of Students 319
For this study, 113 students were selected from a pool of 903 eligible 320 applicants. Only 109 students are part of the study, as 4 did not attend all 321 sessions of their respective course. The gradable component of our online 322 application form focused on essay questions rather than quantitative aspects (i.e. 323 grades). Each application was reviewed and ranked by at least one member of 324 the admissions committee. The admissions committee was blind to previous 325 academic performance and emphasized representation of all corners of the 326 country. Students were required to rank their top 3 courses of a total of 16 327 courses available to them. These included the 6 courses in biology-related fields 328 analyzed for this study, as well as courses in computer science, engineering, 329 entrepreneurship and social sciences. Students were considered part of the 330 "Initial Interest" group if they were assigned to one of the 3 courses they selected. 331 Students were part of the "No Initial Interest" group if they were assigned to a 332 course different from their 3 selections. In total, 32 (29.36%) students were in the 333 "No Initial Interest" group, while 77 (70.64%) students were in the "Initial Interest" 334 group. Instructors were blind to which students were part of each group, although 335 they were aware of the experiment. In order to allow for student preparation, the 336 students were notified of their final course assignment at least 14 days prior to 337 the start of the program. 338 339

Student consent 340
The Clubes de Ciencia Bolivia Review Committee reviewed and approved 341 the design and execution of this experiment and program. In accordance with 342 Bolivian regulations, students were informed about the study prior to the 343 application process. In order to participate in the program, students were 344 required to consent to make pre-and post-test results available for anonymized 345 publication. Upon admission, students had to submit two documents: a copy of 346 an identification document proving their Bolivian citizenship and a letter from their 347 high school or college proving that they were registered students and were in the 348 year they indicated in the application. Occasionally, as this took place during 349 school recess, we also accepted their last school report as proof. In the case of 350 students under the age of 18, legal age in Bolivia, parents had to consent and 351 send a signed and notarized authorization. Because a large number of students 352 resided in different departments, a template letter was provided as part of the 353 admission package. 354 355

Testing 356
In each course, the instructors administered a 5-question test focused on 357 the material to be covered in class lectures. Students were tested at the 358 beginning and end of the course. Tests were graded by both instructors. 359 Importantly, the pre-tests were not returned to the students before the post-test, 360 nor did they have access to the answer key. Moreover, students were unaware 361 that the pre-and post-tests were identical ahead of time. Each test was graded 362 by each of the two instructors of the course. The average of both grades was 363 then computed for analysis. A complete list of questions is available in Table S4. 364

Statistics 366
For analyses that considered two groups of students, Two-tailed paired or 367 unpaired t-tests were performed using PRISM version 7.0, depending on whether 368 the groups being compared contained the same set of students (before vs. after 369 the course) or different students (e.g. No Initial Interest vs. Initial Interest 370 students). Gaussian distributions were assumed. Significance was set at a 371 threshold of p<0.05. 372 For analyses that considered three or more groups of students, we 373 performed one-way ANOVA with Tukey's correction using PRISM version 7.0. 374 Gaussian distributions were assumed. Significance was set at a threshold of 375 p<0.05. 376 To compute the correlation between educational level, type of institution 377 attended, previous academic achievement, or gender and confidence in success 378 in science, we performed a multinomial logistic regression. Bolivian students rate their scientific education highly 386 To gain insight into the lack of relative representation of Bolivia in STEM 387 fields, we conducted a country-wide online survey targeting precollege (11th and 388 12th grades) and college students in the context of an application to a science 389 outreach program (see Methods). Respondents included students from 8 of 9 390 Departments in Bolivia, ages 15-22. A total of 903 students responded to the 391 survey, which included questions designed to gauge participants' perceptions of 392 their educational experience, interest in science, and career goals. Notably, the 393 sample group comprised 48.7% males and 51.3% females, equally spread 394 across public (53.5%) and private (46.5%) institutions. Moreover, 30.0% of 395 respondents were enrolled in high school, while 70.0% were in college (Table  396 S5). 397 Surprisingly, respondents both at the precollege and college levels rated 398 their educational experience highly (7.75±0.08 precollege, n = 268 and 399 7.59±0.05 college n = 614, on a 1-10 scale; p = 0.0736) ( Figure 1A). regardless of education level, gender, or type of institution attended ( Figure 1A). 407 We then asked respondents to indicate what is the highest level of 408 education that they would like to achieve, as well as whether they knew the 409 requirements to pursue a career in science. Curiously, the vast majority (91.8%) 410 of respondents reported to understand the required training to pursue a career in 411 science ( Figure S1A) and hoped to obtain a graduate level of education (77.6% In order to understand if our sample population was biased towards high 419 academic achievers in the sciences, we requested respondents to voluntarily 420 report their academic grade in their previous biology courses. Of the 876 421 respondents who opted to report their grades, 23.3% had never taken a biology 422 course, whereas 25.6% had but achieved a low grade (<80%) and 51.1% 423 obtained a high grade (>80%) ( Figure S1B). Similar numbers were obtained 424 when inquiring about courses in Chemistry, Physics, and Mathematics (Figure 425 S1B). Of note, no strong correlation (r < 0.5) between academic achievement in 426 any of these subjects and perceived education quality was found ( Figure S1C). 427

428
Bolivian students have confidence in their scientific aptitude 429 To gain further insight into students' perceptions and attitudes towards 430 STEM careers, we asked respondents how confident they were in achieving a 431 high score in a STEM subject test. We found the majority of respondents to be 432 confident that they will perform well in STEM courses ( Figures 1C-E). Notably, no 433 clear distinctions were observed in confidence levels when participants were 434 grouped by education level or type of institution attended ( Figure 1C). 435 Nevertheless, confident participants were more likely to be high academic 436 achievers than low academic achievers or have not taken a biology course 437 previously, whereas neutral respondents were the least likely to be high 438 achievers ( Figure 1C). Interestingly, male participants reported higher levels of 439 confidence than their female counterparts ( Figure 1D). Finally, we found no clear 440 correlation (r = 0.1942) between confidence in future STEM subject tests and 441 perception of education quality ( Figure 1E). Altogether, our results show that the 442 Bolivian student population perceives their education as being of high quality, 443 intend to pursue advanced science degrees, and are confident that they will 444 succeed in STEM subjects.  Because the application was open to students from 11th grade in high school to 465 third year of college at the time of application, the student cohort was enriched 466 for college students over high school students ( Figure 2E). Importantly, because 467 the experiment was performed during summer vacation, some students had 468 graduated high school but had not yet started college. Together with the high 469 school students, we categorized this group as "precollege". Moreover, the 470 committee gave preference to applicants who had not attended our program in 471 past years (77 of 109 participants), diminishing the possibility of any advantage 472 gained from previous exposure to our approach that might confound our data 473 ( Figure 2F). In fact, as 2 of 6 courses had been taught in previous years, the 474 returning students were not allowed to attend the same course twice. Altogether, 475 to the best of our possibilities, our study analyzed a cohort of students well 476 balanced with regards to age, gender, previous education quality and level, as 477 well as representative of the demographics of Bolivia. Each course was taught by 478 2 instructors: 1 instructor based in a foreign country (United States or Europe) 479 and 1 local instructor. 480

481
The courses contained a theoretical component, taught using a 482 discussion-based lecture approach ( Figure S2A), as well as a project-based 483 laboratory component ( Figure S2B) [35,36]. At the end of the courses, all 484 students were required to present on a topic within the scope of their course that 485 they researched in small groups of 3-4 students during the week ( Figure S3). We 486 designed questions to assess comprehension of basic principles behind 487 biological processes, such as heredity, immune rejection, or neural 488 communication (Table S4). As we aimed to understand whether students had a 489 basic grasp of these topics, we designed questions targeting knowledge and 490 comprehension, following the principles detailed in the Bloom's taxonomy of 491 questions [37]. Strikingly, we found that incoming students performed poorly in 492 our tests (38.24%±20.42, N = 109) ( Figure 3A), revealing a stark disconnect 493 between perceived education quality and aptitude in STEM subjects. 494

495
In order to interrogate the effectiveness of our approach to teaching 496 biology, we asked whether taking a course increased the score in the same tests. 497 Overall, test scores after taking the courses were significantly higher (N=109, 498 p<0.0001) than those obtained at the beginning of the program ( Figure 3A). In 499 fact, 108 of 109 students increased their test score after participating in the 500 program ( Figure 3A). Of note, this pattern was true for every individual course 501 analyzed ( Figure 3B-G), from the course with the lowest test scores, Genome 502 Engineering ( Figure 3E, n=20 p<0.0001) to the one with the highest scores, 503 Microbiology ( Figure 3F, n=17, p<0.0001), demonstrating that the efficiency of 504 our approach extends beyond a particular topic or instructor. 505 506 Possible confounding factors did not influence students' performance 507 Next, we investigated whether our results were confounded by external 508 factors. Importantly, we found no influence of gender or age on test scores 509 obtained either before or after taking a course ( Figure 4A-B). Curiously, however, 510 students coming from public institutions scored slightly better than those enrolled 511 in private institutions before taking the course (p<0.05), yet no difference was 512 found in scores after taking the course ( Figure 4C). In Bolivia, public universities 513 tend to be higher ranked than their private counterparts [38]. Therefore, the 514 statistically significant difference in scores between students enrolled in public 515 versus private institutions is likely to have been driven by college students, who 516 outnumbered precollege students in our program ( Figure 2E). 517 518 As expected, college students performed better than precollege students 519 in the test before taking the course (p<0.01). Still, no difference was found in test 520 scores after taking a course, indicating that education level did not impact the 521 extent of learning ( Figure 4D). This observation is in agreement with previously 522 published data showing that students from different backgrounds can achieve the 523 same learning gains in project-based biology courses [36]. Moreover, these data 524 are also consistent with the notion that project-based STEM courses "level the 525 field" by preferentially benefiting students at an initial disadvantage [35]. 526 527 One could argue that previous exposure to similar outreach programs 528 could confer an advantage in taking the tests used in this study. As there are no 529 other major non-competitive science outreach programs in Bolivia, as opposed to 530 science olympiads, for example, we only accounted for participations in our 531 initiative in previous years, 2016 and 2017. Comparison of test scores between 532 first time students and those with previous experience revealed no significant 533 difference in test scores either before or after the course between these two 534 groups ( Figure 4E), suggesting that our approach can be used to educate 535 student populations independently of their previous access to similar 536 opportunities. Of note, as mentioned in Methods, returning students were not 537 allowed to retake the same course from previous years. 538 539 Bolivia is a very diverse nation, with large disparities in income and access 540 to education between different departments [24,39]. Abundant literature shows 541 strong correlations between socioeconomic status and educational attainment 542 [40]. Santa Cruz, in particular, is the wealthiest department in Bolivia per capita, 543 and students residing in Santa Cruz are the most represented group in our study 544 (48/109). Nevertheless, our results demonstrate that there is no significant 545 difference in scores either before or after the courses between students residing 546 in Santa Cruz and those coming from other departments ( Figure 4F). However, 547 we did find a modest but statistically significant (p<0.05, N = 96) decrease in the 548 post-course test scores of students who reported to work after school in 549 comparison with those who did not ( Figure S4A). Work obligations from a young 550 age, prevalent in rural parts of Bolivia, have been shown to result in slower 551 learning speeds, disinterest in academics, and ultimately higher school dropout 552 rates [24]. It is thus possible that this small decrease observed in our study hints 553 at differences in learning ability not captured when not specifically sorting out 554 students who work while pursuing their studies. Nonetheless, none of several 555 other possible confounding circumstances impacted test scores after the courses. 556 In addition to those mentioned above, we also found no significant differences 557 after dividing students based on the highest academic degree they desire to 558 acquire ( Figure S4B), or their English skills ( Figure S4C), education level of their 559 parents and siblings ( Figure S4D), perceived education quality ( Figure S4E), 560 previous academic success in biology classes ( Figure S4F), confidence in 561 performance in STEM subject tests ( Figure S4G), previous exposure to STEM 562 professionals ( Figure S4H), or even preferred mode of acquiring new information 563 (visual or verbal; Figure S4I). 564 565 Initial interest in the course material does not affect learning 566 Next, we sought to determine whether the extent of learning depended on 567 interest on the specific topic taught in a course. Interest in a topic is well 568 documented to influence motivation and scholar achievement related to that topic 569 [41]. Conversely, becoming more skilled and knowledgeable about a topic tends 570 to increase an individual's interest on it due to increased positive associations 571 [42]. To investigate the possibility that previous interest in a course altered the 572 extent of learning after participating in that course, we compared test scores 573 between students who were placed in courses outside their list of preferred 574 options (No Initial Interest) with those who attended a course in their preferred 575 field (Initial Interest). Strikingly, students in the "No Initial Interest" group did not 576 perform differently from those enrolled in a course on their list of three top 577 choices (i.e. "Initial Interest" group) across all 6 courses ( Figure 5A-F). Students 578 from either group attained equally significantly higher test scores after taking their 579 respective courses (p<0.0001), suggesting that our model of science teaching 580 was effective even to those students who were not directly interested in the 581 specific subject taught a priori. 582 25 583 In summary, the lack of significant differences in the final, i.e. after taking 584 the course, test scores between genders, age groups, previous exposure to 585 science outreach, education level of the students or their families, previous 586 academic success in biology classes, department of residence, or initial interest 587 in the specific course topic, amongst others, strongly suggest that our approach 588 to teaching biology in a dynamic and inquisitive way may be broadly applicable in 589 other developing countries. 590 591 Hands-on workshops positively impact students' enthusiasm for STEM 592 Finally, we sought to assess the students' satisfaction with our program 593 and characterize their own personal attitudes towards science. When asked to 594 rate different aspects of the CdeCBo initiative upon its completion, the vast 595 majority rated their experience very highly (4 or 5, in a 1-5 scale, with 5 being 596 most satisfied), with regards to having both a foreign and a local instructor 597 teaching together ( Figure 6A), course activities ( Figure 6B) and materials ( Figure  598 6C), motivation of the instructors ( Figure 6D), clarity of course objectives ( Figure  599 6E), and connection between theory and practice ( Figure 6F). These are 600 important data, as the teaching methods utilized by us are vastly different from 601 those used in Bolivia for science education, which could have caused attrition 602 and dissatisfaction amongst students. Furthermore, the students also reported 603 high levels of excitement for science at the end of the program. When asked how 604 strongly they agreed that science is exciting, enjoyable to participate in, a 605 possible career choice, or whether they enjoy solving scientific problems and that 606 hard work will help them succeed in science, most students responded 5 in a 1-5 607 scale, with 5 being strongest agreement ( Figure 7A-E). In additional questions, 608 most students stated they would like to participate in similar programs in the 609 future ( Figure 7F) want to keep learning more about science ( Figure 7H), 610 definitely want to become scientists ( Figure 7I), and thought they could succeed 611 in science ( Figure 7J). More importantly, 89 of 91 students stated that our 612 program increased their interest in science ( Figure 7F). 613

614
To sum up, students in Bolivia demonstrated thorough satisfaction with 615 their education, high confidence in their scientific skills, and a widespread desire 616 to pursue advanced scientific degrees. Even though initial low test scores before 617 taking our courses may have challenged that self-perception, students acquired 618 new knowledge in a teaching methodology new to them, intensive hands-on 619 biology courses, in a manner independent from most possible confounding 620 factors, and reported increased interest in science after completing the courses. Studies in project-based learning have shown that the critical thinking of 636 students exposed to this methodology is significantly increased over time [43]. 637 Yet, not all students transition equally towards project-based learning, and 638 students lacking experience with this approach can experience difficulty in its 639 implementation [44]. Previous studies have shown that this teaching method 640 strongly favors academically "low achievers" [45]. "High achievers", on the other 641 hand, tend to have a difficult transition towards project-based learning, with many 642 of them abandoning the projects altogether [45]. Our student cohort was an 643 interesting group; anecdotally, we observed that our application process recruited 644 academically high achieving students from across Bolivia. The majority of our 645 students have had little, if any, previous experience in project-based learning, as 646 this methodology is not part of the Bolivian curricula. Yet, none of our students 647 abandoned their course and 108 of 109 did better in the post-tests (Figure 3). 648 These results not only challenged our expectations, but also support the 649 importance of performing "on the ground" work across the globe. To date, most small (17-21 students) leads, in our experience, to a close and lasting mentorship 698 relationship between instructors and students, which extends well beyond the 699 timeline of the courses. Together, these characteristics not only powered the 700 efficacious transmission of scientific knowledge, but will also aid in developing 701 and strengthening science in Bolivia in the medium and long term [29]. 702 703 In our opinion, the research universities sponsoring this type of initiatives 704 also benefit from their involvement in tangible ways. Having a graduate student 705 teaching a course abroad allows universities to target two unique pools in 706 recruitment: students in the developed world who are interested in teaching, and 707 students in the developing world who are interested in becoming graduate 708 students in the United States or Europe. 709

710
The current study has some limitations. First, it is not possible to 711 completely account for selection bias in an initiative with eligibility and selection 712 criteria that enrich our sample for students with high academic achievement and 713 an interest in STEM. Nevertheless, we addressed this concern by categorizing 714 students based on their science education perspectives, previous academic 715 achievement, and initial interest in their course, amongst others, and saw no 716 difference in post-course test scores (Figure 4, 5, S4). In addition, comparing 717 students of two different education levels, precollege versus college, revealed no 718 difference in post-course test scores ( Figure 4D). In addition, because the format 719 of the courses included lecture and laboratory components (Table S2) The article goes on to note CdeCBo's lack of emphasis on "ancestral 738 knowledge," and ponders whether CdeCBo's neuroscience course, presenting 739 modern experimentation techniques, should be "reoriented toward primary 740 education, infant health, to students' cognitive processes, or to potentiating these 741 to contribute to development of the country." The following year, following 742 CdeCBo 2018, one print newspaper stressed the need to consider "national 743 reality" in contrast to CdeCBo's vision to "practice a science that is academic, While we fundamentally disagree with the assertion that exposing 747 students to knowledge of modern scientific advancements will negatively impact 748 Bolivia's future, the point is well taken of a need for regional sensitivity and 749 consideration for local contexts. An illustration of CdeCBo's regional 750 considerations comes in topics chosen during course design. Strikingly, some of 751 the above authors performed a study of specific educational desires among 752 Bolivian students in two distinct Bolivian departments, Cochabamba and 753 Chuquisaca, reporting that the most highly sought topic pertained to HIV/AIDS 754 education [54]. Indeed, one of CdeCBo's most popular courses deals with 755 genome editing in the context of immunological resistance to HIV (Table S2). 756 Moreover, a publication from the same group went on to posit that "science clubs 757 [Bolivia] could encourage that the positive image about science and its 758 consequent practice are not contradictory, but continuous… This activity will 759 create a good basis for medical practice to combine assistance with solution, 760 research with intervention." [55]. It is our view that many negative perceptions, 761 among limited segments of population, toward modern scientific education 762 approaches by CdeCBo can be alleviated through enhanced transparency and 763 dialogue between these groups and CdeCBo instructors. We are confident that 764 both camps have in mind Bolivia's best interests for the future, and that open 765 channels of mutually respectful dialogue will identify the most effective and 766 beneficial means of national development. 767

769
To the best of our knowledge, no initiative similar to ours exists yet in the 770 region. Future directions thus include carrying out longitudinal studies to evaluate 771 our students' future success in academic applications and their career choices in 772 STEM versus other fields. In particular, it will be fascinating to investigate 773 whether students who were allocated to courses that did not initially interest them 774 will be any less likely to follow STEM careers than those who attended their 775 preferred courses. Even though test scores were very similar (no statistically 776 significant difference) between these two groups, only a long term follow-up 777 study can shed light on whether their experience overall was similarly inspiring. 778 In either scenario, exposure to biological research during the program has been 779 shown to increase the participants' odds of successfully applying to internships 780 and scholarships to study biology at the college or PhD level [56,57]. Ultimately, 781 we aim to provide an experience which empowers students with the self-782 confidence, excitement, and participatory motivation to help them realize that 783 they too can pursue successful careers as scientists and thrive alongside 784 scientists of any other nation. 785 786

Disclosure Statement 787
The authors reported no potential conflict of interest. level or gender, but it is impacted by type of institution attended (**). Rating 1-10, 806 with 10 being highest education quality. Bars represent mean ± SD. B) Highest 807 educational level desired. C) Confidence in good performance in future STEM 808 subject tests does not depend on educational level or type of institution attended. 809 D) High achievers in biology tests (over 80%) are significantly less likely to be 810 neutral (**) and more likely to be confident (*), compared with low achievers 811 (under 80%) and those who had never taken a biology course. E) Confidence in 812 good performance in future STEM subject tests is affected by gender: female 813 respondents were more likely to be neutral (***), whereas male respondents were 814 more likely to be confident (***). Neither gender was enriched in the group of 815 respondents that felt insecure. F) Correlation between confidence in future good 816 performance in STEM subject tests and perceived education quality (r = 0.1942). 817 Each dot represents the response of a group of students with coordinates 818 (perceived education quality, confidence). Unpaired t-test across groups in (A). 819 Multinomial logistic regression when comparing more than two groups in (C-E). * 820