The Effects of Montessori Education: Evidence from Admission Lotteries Nienke Ruijs TIER WORKING PAPER SERIES TIER WP 14/22 The Effects of Montessori Education: Evidence from Admission Lotteries Nienke Ruijs∗ November 4, 2014 Abstract Little is known about the effects of Montessori education, even though many students in many countries are educated in Montessori classrooms. This study investigates the causal effects of Montessori secondary education by exploiting admission lotteries in Dutch Montessori schools. The results indicate that Montessori education provides an alternative way to attain similar outcomes. Montessori students obtain their secondary school degree without delay at the same rate and with similar grades as non-Montessori students, although the route towards the exams is somewhat different. Further, Montessori students show similar levels of motivation and do not score better on various measures of independence, even though these are the main characteristics Montessori education claims to foster. ∗ The author is affiliated with the University of Amsterdam, the Tinbergen Institute and TIER. Contact: [email protected]. The author would like to thank school A, school B and DUO for providing the data for this study. I am indebted to Hessel Oosterbeek and Sjoerd Karsten for valuable comments and suggestions and help with obtaining the data. Further thanks go to Thomas Buser, Nadine Ketel, Henriette Maassen van den Brink, Noemi Peter, Erik Plug and seminar participants at UvA for helpful comments on earlier drafts. 1 1 Introduction This study investigates the causal effects of Montessori secondary education by exploiting school admission lotteries. Currently, over 22,000 schools in 117 countries are educating their students using the Montessori pedagogy (Association Montessori Internationale, 2013). Notwithstanding the large number of students involved, Montessori methods have passionate advocates and strong opponents. The advocates claim that Montessori education has positive effects on students’ independence and motivation. By providing students with a large amount of choice in their school tasks, independence would be fostered, while motivation is enhanced by minimizing external rewards (Lillard, 2005). Critics, on the other hand, argue that Montessori education can harm academic achievement by a lack of structure and academic standards (Chattin-McNichols, 1992). For parents, students and policy makers, it is important to know which claims are valid. When Montessori education has positive effects on students, schools may be improved by a broader implementation of Montessori methods. Yet, if Montessori education negatively affects students, students and parents could better opt for regular education schools. When there are both positive and negative effects, this is highly relevant information for parents in order to make an informed decision about their child’s education. This paper therefore investigates the effects of being exposed to Montessori education. Despite its popular character, little is known about the effectiveness of Montessori methods. The reason is that students in Montessori schools cannot be directly compared to regular education students. Students and parents select themselves into Montessori education. Parents opting for Montessori schools are, for example, generally of higher socioeconomic status and less often from ethnic minorities (Dohrmann et al., 2007; Rathunde and Csikszentmihalyi, 2005a). While this study is the first to investigate the causal effects of Montessori sec- 2 ondary education, there is a handful of empirical studies for primary education. Lillard and Else-Quest (2006) study students who won and lost an admission lottery for a Montessori primary school in Wisconsin. They find that 5 and 12 year old Montessori students are doing just as well or better than their non-Montessori counterparts on cognitive and social measures. In two older studies, Miller et al. (1975) and Karnes et al. (1970) randomly assigned preschool students to Montessori or non-Montessori Head Start programs, and found no or little initial effects of Montessori education compared to other preschool programs. In the longer run, one study finds higher achievement and IQ for Montessori educated boys, but not for girls (Miller and Bizzell, 1984). The second study finds that Montessori preschoolers more often graduate from high school and experience less grade retention (Karnes et al., 1983). Both studies (and their follow-ups), however, implemented compromised versions of Montessori programs in terms of age grouping, teacher training and day routine and suffer from significant attrition and small samples. This paper circumvents the selection issues described earlier by exploiting school admission lotteries. The admission lotteries basically create a series of randomized experiments: they determine by chance which students are educated at Montessori schools and thereby create valid treatment and control groups. By using administrative data on academic achievement and questionnaire data on socio-emotional functioning, this study investigates a wide range of potential outcomes of Montessori education. The results show that the academic achievement of Montessori students is akin to the academic achievement of students in regular secondary education. Montessori students obtain their secondary school degree without delay at the same rate and with similar grades as non-Montessori students. The socio-emotional functioning of Montessori and non-Montessori students turns out to be comparable as well. Students show similar levels of motivation and Montessori students do not score better on measures of independence. These results imply that neither the claims of 3 the opponents, nor the claims of the critics are supported by the evidence. The paper is structured as follows. Section 2 discusses the characteristics of Montessori education. Section 3 provides additional background information on secondary education in the Netherlands. Section 4 and 5 describe the data and empirical strategy. Section 6 presents and discusses the results and Section 7 concludes. 2 Montessori education 2.1 Montessori primary education Montessori education was developed by Maria Montessori, starting with the ’Casa dei Bambini’ in a poor neighborhood in Rome in 1907, and spreading over the world ever since (Whitescarver and Cossentino, 2008). As Montessori education is most renown for primary education, this section starts with a description of Montessori primary schools. When walking into a Montessori classroom, there are several features that distinguish it from traditional primary education classes. While students in traditional education mainly learn from texts, students in Montessori classrooms are often working with their hands using Montessori materials (Lillard, 2005). Montessori materials are largely self-correcting educational objects designed to teach subjects and concepts via repeated use and in a structured sequence. To teach writing skills, for example, pupils practice the phonetic sounds of letters while tracing letters made from sandpaper with their fingers. In the same period, they develop the motor skills needed for writing with drawing tasks using metal frames. Later, the two activities come together (Lillard, 2005). Other distinctive features of Montessori classrooms are that classes are organized in three-year-age groupings, such that younger students can learn from older students (Montessori, 1972). Students are allowed to choose their own activities 4 from a young age onwards (Mooney, 2000) and generally work and get instructed on their own or in small groups (Whitescarver and Cossentino, 2008). Further, there is an absence of formal grades and tests. One of the main goals of Montessori education is to help students to become independent. Maria Montessori believed that children need to acquire independence in order to grow and develop (Montessori, 1989b). In early childhood education, for example, independence is fostered through practical life exercises such as table washing (Lillard, 2005). Moreover, children in Montessori classrooms can freely choose their work, which gives them experience in making choices (Montessori, 1989a). Another aim of the changes in classroom environment is to create an environment where students are intrinsically motivated to learn. The Montessori materials are designed to sparkle children’s interest. Students actively manipulate the objects and learn through experience, which is claimed to enhance learning and motivation (Lillard, 2005). Increasing motivation is also the argument for the absence of formal grades and tests. Traditional schools offer constant feedback on achievement by providing grades and rewards, thereby replacing intrinsic motivation towards learning by the need for external rewards. Maria Montessori believed that such extrinsic rewards are not necessary and even disrupt students’ learning (Lillard, 2005). 2.2 Montessori secondary education While Montessori education is often associated with primary education, it also has a lively and international secondary education community. An important part of the history of Montessori secondary education lies in the Netherlands. Just 7 years after the opening of the ’Casa dei Bambini’, the first Dutch Montessori primary school opened. As soon as the first students finished primary school, demand for a Montessori secondary school developed. In 1928, the first Montessori secondary school in the world was opened in the Netherlands (Calff, 1980). Today, the school 5 is still in existence and it is one of the schools considered in this study.1 Montessori secondary schools adhere to the famous Montessori adagio: “help me to do it alone”. Similar to primary education, Montessori methods in secondary education are organized towards fostering independence (Montessori, 1973) and creating an environment for intrinsic motivation (Rathunde and Csikszentmihalyi, 2005b). Modern2 Montessori secondary schools differ from regular schools in a number of ways. Most salient is the amount of choice students are allowed in their schoolwork: part of the hours at school are specifically allocated for free choice of activities (Rathunde and Csikszentmihalyi, 2005a). Students are allowed to choose which teacher they want to join during these hours. Further, students can choose when they want to take tests in the lower grades of secondary education3 and Montessori secondary schools focus more broadly than academic achievement. Field projects, such as internships, are organized to promote social development and to learn students how to function in society (Seldin and Epstein, 2003; Rubinstein, 2008). 3 Secondary education in the Netherlands Dutch students and their parents have free school choice, they are not restricted by catchment areas or school fees. Moreover, virtually all primary and secondary schools are completely publicly funded. This includes the religious and special program schools, such as Montessori schools. The government funding is nationally determined and largely dependent on student numbers. Schools can get additional 1 Nowadays there are 16 Montessori secondary schools in the Netherlands. All schools participate in the Dutch national exams, so the content of the curriculum has large overlap with regular secondary education. The Dutch Montessori secondary schools conduct visitations amongst each other to guard the Montessori quality of the schools. 2 As Maria Montessori passed away while developing Montessori methods for adolescents Lillard (2013), Montessori secondary education is largely developed by the Montessori community. Maria Montessori’s writings on secondary education include a radical boarding school program called ’Erdkinder’. This program has been implemented very sparsely, and is not the focus of this paper. 3 At some schools, students can also determine the timing of (part of) the tests in the higher grades of secondary school. 6 Figure 1: The Dutch secondary school system funding for students from disadvantaged neighborhoods. Dutch secondary education starts at age 12, and lasts four to six years. The school system is highly tracked, Figure 1 gives a graphic description. Not all secondary schools offer all school tracks and schools can educate children of different tracks together. The highest track (pre-university education, vwo) takes six years, and gives access to university education. The intermediate track (senior general secondary education, havo) takes five years, and gives access to universities of applied sciences, also called higher professional education. The lowest tracks (pre-vocational education, vmbo) last four years, and give access to vocational education programs. Within the pre-vocational track, there are four different levels. Only the highest pre-vocational level (tl, from hereon referred to as pre-vocational IV) and the senior general secondary and pre-university tracks are offered at the Montessori schools in this study. The initial school track is based on the school track advice of the primary school teacher and on a standardized high stakes test (in most cases a nationwide exit test called the “citotoets”) at the end of primary school. Dependent on student achievement and school policies, students can change track during secondary education. 7 Further, students can follow a higher track after finishing a lower track.4 Subject to some conditions, students can choose their own courses in the second half of secondary education. Secondary schools have to follow national curriculum guidelines, and all students take centrally determined national exams at the end of secondary school. The national exams count for 50% of the final grade, the other 50% is determined by school specific exams taken in the last two or three years of secondary education. 4 Data 4.1 Admission lotteries in Montessori schools In this paper, data from school admission lotteries in two Dutch Montessori secondary schools are used. The names of the schools are anonymized to school A and school B. The schools generally receive more applications than they can enroll and conduct lotteries to allocate the available places. In both schools, the lotteries are organized for each school year separately and are executed by a notary. School A conducts school admission lotteries since 2003. The lotteries are divided into a lottery for students of the two highest school tracks (senior general secondary/pre-university) and for the highest pre-vocational track (pre-vocational IV). Not all applying students participate in the lottery. Students with a brother or sister at the school, children from employees and students applying from a Montessori primary school get placed with priority.5 Table 1 shows the number of lottery participating students and the number of priority students per year and school 4 In that case, students enter the higher track in the year before the final exams. When doing pre-university education after senior general secondary education, for example, students enter in year 5 of pre-university education. When graduating at once, they have their pre-university degree after 7 years of secondary education instead of 6. 5 Since the majority of the lottery participating students subscribed to school A, this implies that the results are largely identified on a sample of students who did not attend Montessori primary schools. Section 6.4 describes this issue in more detail. Of all students receiving priority to school A, 90.2% gets priority because they were educated at a Montessori primary school. 8 Table 1: Distribution of lottery participating and priority students over years School year Lottery winner Lottery loser Priority Total School A: Senior general secondary/pre-university education 2003/2004 50 46 169 265 2004/2005 16 59 203 278 2005/2006 47 9 196 252 2009/2010 37 42 163 242 2010/2011 29 17 178 224 Total 179 173 909 1261 School A: Pre-vocational IV education 2003/2004 17 45 53 115 2004/2005 10 30 60 100 2007/2008 10 29 62 101 2008/2009 29 3 42 74 2009/2010 5 10 43 58 Total 71 117 260 448 School B: Senior general secondary/pre-university education 2008/2009 52 15 60 127 2009/2010 83 4 16 103 2010/2011 38 9 43 90 2011/2012 56 3 34 93 Total 229 31 153 413 Grand Total 479 321 1322 2122 Note: This table reports the number of lottery participating students and the number of priority students per school, year and school track. track. In total, 540 students participated in a school admission lottery for school A.6 In school B, students from Montessori primary schools do not get priority (brothers and sisters of current students and children of employees do). Instead, the lottery is based on residence: the school proportionally takes up students from 3 regions. When there is over-subscription from one region, students from this region partici6 The actual number of students applying to school A for years and school tracks that conducted lotteries is slightly higher: 1759 instead of 1709. However, 44 students (2.5%) could not be merged at the Dutch educational administration. All of these students lost the lottery, mainly in the earlier years, but there is no selection on gender, citoscore and school advice. Other background characteristics are not available for these students. Further, 6 students (0.3%) are dropped because of missing information on ethnicity or living in a disadvantaged neighborhood. 9 pate in a lottery. School B does not offer the pre-vocational IV track. For school B, 260 students participated in a school admission lottery, bringing the total number of lottery participating students to 800.7 Since the effects of Montessori education are studied by comparing students winning admission lotteries to students who lose the lottery and therefore attend another school, it is important to know the characteristics of the alternative schools. The majority of the lottery losing students (80.1%) are enrolled in schools with regular education programs. 8.5%8 still attend Montessori schools and 11.4% attend a Dalton school, a school with a pedagogy comparable to Montessori. With respect to school tracks, most lottery losing students are enrolled in schools similar to school A and school B. 0.63% of the lottery losing students attend schools that only offer the pre-university school track, while 22.8% go to schools that also offer the lower pre-vocational education tracks or attend schools that do not offer the highest track(s). Table 2 describes differences in school characteristics between school A and school B and their alternative schools attended by lottery losers in the first year of secondary education. In general, the standard deviations of the average school characteristics of the alternative schools are large, especially for school A, indicating that lottery losing students attend a wide range of alternative schools. For most characteristics, the alternative schools are significantly different from school A and B. The first seven rows of Table 2 pertain to school quality measures of the Dutch education inspectorate. The patterns indicate that the Montessori schools have less delay in the higher grades and higher average exam scores, although grade progression in the lower grades is lower. It is important to note that these differences 7 202 students from regions without lotteries in a certain year are dropped from the sample. Comparable students at school A, in years for which a certain school track did not have a lottery, are not available for analysis. 1 student (0.2%) could not be merged at the Dutch educational administration. 20 students (4.6%, of whom 13 participated in the lottery) are dropped from the sample because of missing information on primary school advice. 8 17 losing students (5.4%) are admitted to School A or school B from a waiting list, details are in Section 5. The other 10 students (3.2%) attend another Montessori secondary school. 10 cannot be interpreted as effects of Montessori education, as they reflect average school characteristics that are potentially created by differences in school populations. The four bottom rows report more general school characteristics. It turns out that school A is significantly larger and school B is significantly smaller than their alternative schools. With respect to the percentage of students from disadvantaged neighborhoods, school A has more students from disadvantaged neighborhoods in the pre-vocational III/IV and pre-university tracks, but fewer in the senior general secondary track. There are no significant differences in the percentage of students from disadvantaged neighborhoods for school B compared to the alternative schools attended by its lottery losers. 4.2 9 Data on academic achievement The information on students’ academic achievement is provided by the Dutch educational administration (Dienst Uitvoering Onderwijs, DUO). The most recent data available are exam scores of the exams administered in May 2012 and educational positions in October 2012. Students’ final exam results are used to create an indicator that equals one if the student obtained the degree of the advised school track (or higher) on time. Otherwise the indicator equals zero. Simple exam grades are not a good outcome measure, since students can change school track during secondary school and exams are made for each school track separately. When a student changes to a lower track, it is likely that his grades will be higher than they would have been at the higher school track. Therefore, the school track advice of the primary school teacher is used as the school track the student is expected to obtain by the end of secondary school. This advice is compared to the obtained 9 The high percentage of students from disadvantaged neighborhoods is somewhat misleading for school A and its alternatives. It is based on a policy taking into account the characteristics of students’ residential neighborhoods. In the city of school A, this caused some peculiarities in which ’elite’ schools were suddenly identified as schools with many disadvantaged students. The old policy based on individual student characteristics was still in effect in the first years considered in this study. According to that definition, the percentage of disadvantaged students was nearly zero for school A, and around 4% for its alternatives. 11 Table 2: Characteristics of alternative schools attended by lottery losers School A Index grade progression lower grades School B Mean Mean (SD) Mean Mean (SD) school A alternatives school B alternatives (1) (2) (3) (4) 94.07 100.96*** 99.03 101.08** (13.79) % no delay pre-voc. III/IV 81.19 (5.05) 74.70*** (10.68) % no delay senior gen. 67.55 45.65*** 75.26 (13.47) % no delay pre-uni. 67.57 51.06*** (4.91) 60.66 (15.07) Mean grade pre-voc. III/IV 6.36 62.19*** 68.27*** (9.01) 6.13*** (0.25) Mean grade senior gen. 6.44 6.03*** 6.52 (0.26) Mean grade pre-uni. 6.57 6.08*** (0.13) 6.62 (0.33) Number of students 1579.97 823.49*** 61.83 6.35*** (0.15) 788.02 (304.67) % disadvantaged pre-voc. III/IV 6.16*** 1079.96*** (301.03) 33.43*** (39.18) % disadvantaged senior gen. 53.15 60.28** 7.31 (24.74) % disadvantaged pre-uni. 56.21 50.49** 6.97 (14.34) 3.49 4.73 (21.51) (12.40) Note: Columns (1) and (3) report school characteristics of school A and school B. Columns (2) and (4) report the means and standard deviations of each characteristic for the alternative schools attended by lottery losers in the first year of secondary education for school A and B separately. The information in the table is derived from the Quality Cards 2002-2010 of the Dutch education inspectorate. The education inspectorate publishes secondary school information on a yearly basis. The measures published in the year of school choice are used for each lottery participating student. This creates an average weighted by the number of winning students in each year for the Montessori schools. For the alternative schools, this creates an average weighted by the number of attending lottery losers in each year. The first seven rows describe school quality measures. The index for grade progression in the lower grades is a measure of the percentage of students getting to the third grade without grade retention, correcting for students who attend higher or lower school tracks than their primary school advice. The other measures pertain to the higher grades, and report on the percentage of students passing through the higher grades without delay and the average grade on the final exams. The stars refer to the p-values of one-sample t-tests testing whether the means of the alternative schools equal the means for school A or school B, where the means of school A and B are the constants. ***p<0.01 **p<0.05 *p<0.10. 12 degree. When a student gets a degree from a lower track or when he is grade retained, he is coded as not obtaining his degree on time. When a student follows the advised or a higher school track and obtains his degree without delay, he is counted as obtaining his degree on time.10 For students with combined school track advices (e.g. senior general secondary/pre-university education) delays or lower degrees are counted from the lowest of the two tracks.11 To study whether Montessori education has an effect on gaining high grades, an indicator variable for obtaining a degree on time with at least a grade point average of 7.5 (on a 10 point scale) is used. Further, indicators for obtaining a degree on time with at least a passing grade for Dutch, English and mathematics are constructed.12 Since the youngest cohort of students started secondary education in 2011/2012, exam outcomes are only available for 47.9% of the lottery participating students in the sample. To investigate academic outcomes for a larger group of students, I consider students’ educational positions measured four years after participating in an admission lottery. The fourth year is chosen because students are clearly tracked by the time they are in 4th grade.13 In lower grades, students can be educated in combined classes, such as senior general secondary/pre-university education. Two different outcome measures are used for the fourth year. The first one is 10 It can also be that a student gets a higher degree with some delay. In that case, he is coded as obtaining his degree on time when he gets the higher degree at the same time as he could have gotten that degree if he first followed the school track of his primary school advice. For example, a student with a senior general secondary education advice is considered to obtain his degree on time if he passes his pre-university exams in seven years. 11 Counting from the highest of the two tracks does not change the results for the exam measures. For the 4th year, there is a slight difference, indicating that students at Montessori schools are more often at the lower track of the combined advice in 4th grade. 12 To ensure that grades are comparable across schools, the national exam grades are used. A grade of 5.5 or higher (on a 10 point scale) is considered a passing grade. It is possible for Dutch students to fail courses and still obtain their degree. Dutch secondary education has different levels of mathematics within the same school track. This distinction is not taken into account in the mathematics variable. Alternative specifications (computing the indicator for each math level separately and correcting math grades for difficulty following Leuven et al. (2010)) yield similar results. 13 In this paper, 4th grade refers to the 4th grade of Dutch secondary education, in which students are approximately 16 years old. 13 whether a student is in 4th grade at the advised school track or higher after four years of secondary education. Basically, this measure captures the students who are on track to obtain their degree on time: grade retained students are coded as zeros. Since parents and students might care more about the school track a student eventually obtains instead of whether he is grade retained or not, grade retainers are not taken into account in the second measure. This measure indicates whether students who are in 4th grade after four years are educated at their advised school track or higher. As described in Section 3, students can choose a course program in the second half of secondary education. The science program is considered the more difficult program in the senior general secondary and pre-university school tracks (Buser et al., 2014). Since (not) choosing a science program strongly affects the options for tertiary education, it is interesting to see whether Montessori students are more or less often enrolled in the science program in the higher grades of the academic school tracks. 4.3 Questionnaire on socio-emotional functioning To measure the effects of Montessori education on socio-emotional functioning, a questionnaire was used. The content of the questionnaire is described in Table 3. The table includes information on the instruments used to measure each outcome, the reliability of these instruments and sample questions. As independence is one of the main characteristics that Montessori education is claimed to foster, the questionnaire includes two measures of independence. Independence at school measures the degree to which students are able to effectively work on school tasks with classmates or on their own. As no existing instrument was available, it was developed for this study. The scale includes questions on taking initiative, working together and evaluating the learning process. Independence out of school focusses more broadly on functioning as an autonomous individual. 14 15 Developed for this study Interest in society iii 7 16 8 23 22 questions Number of I consider it very important to vote at elections I get along well with most of my teachers I would like to learn a lot at school I think about how my actions will affect others an answer myself, before turning to the teacher If there is something I don’t understand, I first look for Sample question 0.58 0.84 0.83 0.75 0.68 in sample Cronbach’s α Parental support Developed for this study 3 My parents often help me with my homework 0.57 Note: Column 5 reports Cronbach’s α statistics which are computed using the students who responded to the questionnaire. Cronbach’s α is a measure for reliability. It is a coefficient of internal consistency, indicating the extent to which items on the same scale are measuring one common characteristic.i Some items adapted from L&S (Vorst, 1993), ii Beckert (2007), iii Anderson et al. (1994), iv Smits (2008). SVL 2008: PS, RL Well-being at school iv SVL 2008: LG iv CASE: ET, CV , WAS: BA Motivation towards learning tasks Independence out of school ii Developed for this study Independence at school i Instrument Outcome Table 3: Content and reliability of the questionnaire It includes sub-scales from existing questionnaires on evaluative thinking, need for confirmation (comparative validation) and behavioral autonomy. To study whether Montessori students are indeed more motivated for school, a scale on motivation towards learning tasks is included. To get a broader view on students’ socio-emotional functioning, the questionnaire also includes questions on well-being at school, interest in society and on problematic behavior like truancy and use of alcohol and drugs. Further, the questionnaire includes items on parental support to school matters and participating in commercial school support to investigate whether selective parental compensation diminishes potential differences in academic achievement. For budgetary reasons, a sample of 910 students was invited to fill out the questionnaire. The details of the sampling procedure are described in Appendix A. The students received a letter on university stationery, explaining that they applied for a Montessori secondary school some time ago and that the University of Amsterdam is curious to see how their secondary school time is going or went. The letter contained a reference to a website, a personal login code and password, and students received a reward of €10 when they completed the questionnaire. 609 students (67%), including 415 lottery participators, responded to the questionnaire. Table A2 in Appendix A describes the characteristics of students who did and did not respond to the questionnaire. It turns out that nonrespondents are older, more often lottery losers and have lower school advices. Further, they are more often males and non-western migrants. While it is unfortunate that there is selective nonresponse, the pattern of nonresponse is not uncommon (e.g. Van Loon et al., 2003; Søgaard et al., 2004). Moreover, the balancing tests reported in Table 4 indicate that nonresponse does not create differences between lottery winners and lottery losers in terms of background characteristics. 16 5 Empirical strategy The effects of Montessori education are hard to identify, since the choice for a Montessori school is endogenous. Different parents choose for Montessori and regular education, and parental characteristics determining Montessori choice are likely to be related to student outcomes. The admission lotteries to school A and B, however, basically create a series of randomized experiments on Montessori school attendance. To investigate the effects of Montessori education, lottery participating students attending and not attending school A and B are compared using: 0 yil = Xil β + γM ON Til + νl + εil (1) where yil indicates outcomes y of student i participating in lottery l. Xil0 is a vector of student characteristics added for precision and includes gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. ν l is a fixed effect for lottery l, capturing the different lotteries in different years and schools. When a school has multiple lotteries in a single year (for different school tracks or regions), a lottery fixed effect is included for each lottery. The lottery fixed effects make sure that students within the same lottery are compared, instead of comparing students in lotteries with potentially slightly different characteristics. Since the lotteries are year and school specific, there are no year or school fixed effects. εil is the error term. M ON Til is an indicator variable for ever attending Montessori school A or B, making γ the parameter of interest. Even in the sample of lottery participating students, ever attending a Montessori school may be endogenous. Over time, some students who initially lose the lottery are offered a place on the Montessori schools. To be specific, 43 students are present 17 at school A or B in some point of time after they lost the lottery, of whom 17 attend the schools from the first year onwards, the always-takers.14 These students were on the waiting list, and were placed when a slot came free. Not all students on the waiting list decide to go to the Montessori schools when a slot comes free, some decide to stay at their alternative secondary school. This choice may depend on student characteristics, making the indicator for ever attending a Montessori school potentially endogenous, which would bias γ. Therefore, an instrumental variables framework is used, using the outcome of the lottery as an instrument for ever attending a Montessori school. The first stage is defined as: M ON Til = Xil0 π + λW IN il + υl + ηil (2) where W IN il is an indicator for whether student i is a lottery winner. In this framework, the instrument, W IN il , should be both relevant and valid. For the first condition, winning the lottery should predict Montessori school attendance. Appendix B shows first stage results confirming that the lottery is indeed a relevant instrument. For the instrument to be valid, it should be uncorrelated with the error term εil . It is assumed that winning the lottery has no effect on student outcomes except from the effect via Montessori school attendance. This implies, for example, that disappointment about losing the lottery should not lead to a decrease in motivation for school. Further, the outcome of the lottery should be randomly determined, which is very likely since the lotteries are executed by a notary. To show that the lotteries are indeed conducted fairly, Table 4 shows balancing tests testing whether there are differences in observable characteristics between lottery winners and lottery losers. Columns 1 and 2 show the means and standard deviations for students losing and winning a school admission lottery. Columns 3 and 4 14 Reversely, no never-takers are found: all students who won the lottery are present at the Montessori schools for at least one year. 18 show the actual balancing tests, regressing the dependent variables denoted in each row on an indicator for winning the school admission lottery. The balancing tests show that for nearly all characteristics, there are no significant differences between winning and losing students. The exception is age: winning students are somewhat older than losing students, although the coefficient is small. Table 4 reports separate regression coefficients to facilitate interpretation. To test whether the coefficients are jointly significantly different from zero, a joint balancing test was conducted. Regressing the indicator for winning the lottery on all background characteristics, it turns out that the null hypothesis cannot be rejected (F(10,774)=0.781, p=0.647), supporting the conclusion that the lotteries were indeed conducted fairly. Since not every lottery participating student participated in the questionnaire, columns 5 and 6 show similar balancing tests for winning and losing students who responded to the questionnaire. The results are comparable to the results of the previous balancing tests15 , indicating that winning and losing students who responded to the questionnaire are similar in terms of background characteristics. 6 Results 6.1 The effects of Montessori education on academic outcomes The results in Table 5 indicate that there is little impact of Montessori education on students’ academic achievement. The first row shows that after four years of secondary education, 75.2% of the students is in 4th grade of the advised track or better. On this outcome, there is no significant difference between Montessori and non-Montessori students. When a student is not in the advised track in 4th grade 15 The null hypothesis of the joint balancing test cannot be rejected either (F(10,389)=1.001, p=0.441). 19 20 (Table continues on next page) Western migrant Non-western migrant Dutch Boy primary school advice Pre-uni. primary school advice Senior gen./pre-uni. primary school advice Senior gen. primary school advice Pre-voc. IV/senior gen. primary school advice Pre-voc. IV Citoscore variable Dependent 0.102 (0.303) (0.308) (0.338) (0.407) 0.106 0.132 (0.424) (0.465) 0.209 0.766 (0.483) (0.465) 0.685 0.367 (0.452) 0.286 (0.462) 0.307 (0.445) 0.271 (0.223) 0.052 (0.277) 0.315 (0.405) 0.206 (0.418) 0.224 (0.422) 0.231 (0.312) 0.109 (0.422) 0.084 (4.380) (5.229) 0.231 541.517 (2) (1) 540.438 lottery winners lottery losers Mean (SD) for (0.026) 0.019 (0.030) -0.003 (0.037) -0.016 (0.042) 0.028 (0.034) -0.012 (0.036) 0.008 (0.038) 0.005 (0.019) -0.011 (0.020) 0.010 (0.293) 0.240 (3) tests lottery Balancing 0.450 0.914 0.662 0.504 0.730 0.825 0.900 0.567 0.620 0.413 (4) P-values Table 4: Descriptive statistics and lottery balance Mean (SD) for (0.031) 0.008 (0.037) 0.047 (0.045) -0.055 (0.052) 0.016 (0.044) -0.026 (0.045) -0.032 (0.047) 0.067 (0.024) -0.015 (0.025) 0.006 (0.376) 0.217 (5) respondents questionnaire lottery for Balancing tests 0.802 0.200 0.218 0.756 0.564 0.482 0.152 0.527 0.805 0.565 (6) P-values 21 (0.499) 0.542 (0.446) (0.465) 0.315 (0.415) 12.109 (2) (1) 12.087 lottery winners Mean (SD) for lottery losers Mean (SD) for (0.039) -0.023 (0.040) 0.079 (3) tests lottery Balancing 0.559 0.047 (4) P-values (0.049) -0.020 (0.048) 0.082 (5) respondents questionnaire lottery for Balancing tests 0.683 0.085 (6) P-values Number of students 321 479 800 415 Note: Columns (1) and (2) display the means and standard deviations for students losing and winning a lottery for a Montessori school. Columns (3) and (4) report separate regression coefficients and the p-values of the variables indicated in each row on an indicator variable equalling 0 if the student lost the lottery and equalling 1 if the student won the lottery. The sample is restricted to students participating on the lottery. Columns (5) and (6) report similar regression coefficients for the sample of lottery participating students who responded to the questionnaire. All regressions include lottery fixed effects. Robust standard errors are reported in parentheses. neighborhood Disadvantaged Age at application variable Dependent Table 4: (continued) after four years of secondary education, this can have two different causes: it can be that the student is grade retained, or that he has moved to a lower school track. The second line of results indicates that Montessori and non-Montessori students differ in these causes: when grade retained students are dropped from the sample, it turns out that Montessori students who are in 4th grade after four years more often attend a lower school track than can be expected from their primary school advice. This indicates that students at the Montessori schools are more often sent to a lower school track, while similar students in traditional schools are grade retained.16 More important than fourth year outcomes are the final exam results. It turns out that Montessori students obtain the degree of their advised school track on time at the same rate as non-Montessori students, which matches the results on the first row of Table 5. Since Montessori students are more often educated at lower school tracks in 4th grade, one might expect fewer Montessori students to obtain the degree of their advised school track with one year delay: students at a lower track may finalize their secondary education at that lower school track, while grade retainers obtain the degree of their initial track with one year delay. This conjecture is not seconded by the results: Montessori education has no significant effect on obtaining the degree of the advised track with one year delay. It might be that Montessori students catch up by doing the higher school track after their first exams, which also yields a degree with one year delay. With respect to grades, it turns out that Montessori education has no significant impact on obtaining a degree with high average grades, nor on obtaining a degree with passing grades for Dutch, English and Math. Further, Montessori education 16 Both Montessori schools indeed have an avoiding policy towards grade retention. Another potential explanation is that students who lose the lottery go to schools that only offer higher school tracks. To avoid changing schools, students at these schools might prefer grade retention over moving down a track. Indeed, 41.8% of the lottery losing students at school A attend a school that does not offer the pre-vocational IV school track. School B, however, only offers the two highest tracks, and 77.4% of their lottery losing students attend schools with the same school tracks, while 22.6% attend schools that also offer lower, pre-vocational, tracks, implying that this mechanism should not operate in school B. As shown in Table C1, the fourth year outcomes are not significantly different for school A and school B, suggesting that school policies are more relevant than avoiding to change schools. 22 Table 5: Results on academic outcomes Outcome In 4th grade of advised track or better after four years Mean (SD) N (1) (2) 0.752 625 -0.049 -0.022 (0.048) (0.047) -0.079** -0.080** (0.038) (0.037) -0.049 0.000 (0.064) (0.060) -0.017 0.008 (0.058) (0.057) -0.035 -0.036 (0.030) (0.030) -0.047 -0.015 (0.069) (0.068) 0.022 0.057 (0.069) (0.068) -0.035 0.029 (0.078) (0.075) -0.069 -0.030 (0.053) (0.052) 0.441 0.406 (0.302) (0.308) 0.115* 0.112* (0.064) (0.066) (0.432) In 4th grade of advised track or better 0.895 after four years (cond. on being in 4th grade) Obtained degree on time (0.307) 0.671 on advised track or higher 525 383 (0.470) Obtained degree on advised track or higher 0.777 with at most one year delay 376 (0.417) Obtained degree on time on advised track 0.047 with at least a 7.5 on average 381 (0.212) Obtained degree on time on advised track 0.517 with passing grade for Dutch 377 (0.500) Obtained degree on time on advised track 0.561 with passing grade for English 378 (0.497) Obtained degree on time on advised track 0.500 with passing grade for Math 308 (0.501) Enrolled in science program 0.184 483 (0.388) Parental support 8.664 408 (2.406) Participated in commercial 0.491 school support (0.501) 407 ! Controls Note: Each row reports two IV regressions with winning the lottery as an IV for ever attending a Montessori school. The sample is restricted to students participating on the lottery. The first column displays the means and standard deviations for these lottery participating students. Columns (1) and (2) report IV estimates without and with controls. Robust standard errors are reported in parentheses. Alternative specifications of the standard errors (e.g. clustering on school attended in the first year or clustering on school attended in the first year*school advise*year) yield similar results. All regressions include lottery fixed effects. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. Exams are graded on a 10-point scale, where grades above 5.5 are considered a pass. ***p<0.01 **p<0.05 *p<0.10. 23 does not affect the probability of being enrolled in a science program. It is conceivable that the limited differences in academic outcomes in either direction are created by selective parental compensation. Parents might send badly achieving students to commercial school support activities (such as private tutoring) or decide to invest more time themselves, while they do not interfere when a student is doing well. The results in Table 5 show that there is no significant difference in the amount of parental support between Montessori and non-Montessori students. Montessori students do, however, receive more commercial school support, although the coefficient is only significant at the 10% level. An alternative explanation is that the coefficients are imprecisely estimated. Looking at the signs of the coefficients, it should be noted that all coefficients on additional support are positive, while 13 out of the 18 coefficients for the academic outcomes are negative. Since the standard errors are rather large, the possibility of substantial negative effects cannot be excluded. Overall, the results indicate that Montessori students achieve just as well as students attending regular secondary education. 6.2 The effects of Montessori education on socio-emotional outcomes While the previous section focussed on academic achievement, the results in Table 6 focus on socio-emotional functioning. Even though Montessori education aims to improve students’ independence and motivation, the results show no positive effects of Montessori education on these outcomes. There are no significant differences between Montessori and non-Montessori students on the measures of independence at school and motivation towards learning tasks. Contrary to the expectations, Montessori education has a negative impact on independence out of school. Exploring the sub-scales (results not shown), it turns out that this result is mainly caused by the sub-scale of evaluative thinking. Evaluative thinking measures the extent to 24 which students think about their actions and decisions before acting. On the other measures of socio-emotional functioning, there are few significant effects of Montessori education. The exception is well-being at school, which indicates that Montessori students enjoy their school better and have a better relationship with their teachers. Montessori education does not significantly affect problematic behavior nor students’ interest in society. Including the academic outcomes, 20 different variables were tested in the analyses above. At a 5% significance level, you would expect one variable to turn out significantly by chance: a Type I error. When applying the Bonferroni and the Bonferroni-Holm method to control for multiple testing, none of the null hypotheses can be rejected. Even though the Bonferroni and Bonferroni-Holm methods are very conservative, the results should be interpreted with caution. The main message to take out of the results is that the outcomes of Montessori students are similar to the outcomes of non-Montessori students. 6.3 Heterogeneous effects The results presented so far are estimates based on two Montessori secondary schools. Although both schools are Montessori schools, it might be that one school attains better outcomes than the other. Table C1 in Appendix C shows heterogeneous effects for school A and school B on the main outcomes considered in this study.17 It turns out that the effects on academic achievement after four years of secondary education are not significantly different between the two schools.18 The finding that Montessori students more often participate in commercial school support, however, is driven by students at school B. 17 The interaction effects for the other outcome measures (on grades and problematic behavior) are insignificant for school A and school B and for boys and girls. For the different school tracks, it turns out that compared to senior general secondary/pre-university students, the pre-vocational IV students in Montessori education more often obtain a passing grade on English and Math. Further, these students are less often late, truant or sent out of class. 18 Differential effects on exam measures are not available because the students participating in admission lotteries for school B are too young for secondary school exams. 25 Table 6: Results on socio-emotional outcomes Outcome Mean (SD) N (1) (2) 73.520 408 -0.008 0.095 (0.917) (0.938) -2.253** -1.923** (0.975) (0.974) -0.369 -0.269 (0.552) (0.555) 1.302 1.462* (0.823) (0.830) 0.770 0.660 (0.484) (0.506) -3.022 -5.833 (4.019) (4.297) 0.084 0.082 (0.052) (0.054) 0.048 0.014 (0.057) (0.058) 0.041 0.024 (0.054) (0.051) Independence at school (7.619) Independence out of school 83.475 408 (7.936) Motivation towards learning tasks 17.895 409 (4.494) Well-being at school 37.875 409 (6.570) Interest in society 23.627 408 (3.994) Number of times being late, truant or sent 21.593 out of class in the past 6 months 407 (39.622) Smoked cigarettes or drank alcohol 0.597 in the past month 407 (0.491) Used drugs of any kind 0.373 in the past 12 months 407 (0.484) Ever got a fine or was suspected of a 0.270 criminal offense by the police (0.445) 407 ! Controls Note: Each row reports two IV regressions with winning the lottery as an IV for ever attending a Montessori school. The sample is restricted to students participating on the lottery. The first column displays the means and standard deviations for these lottery participating students. Columns (1) and (2) report IV estimates without and with controls. Robust standard errors are reported in parentheses. Alternative specifications of the standard errors (e.g. clustering on school attended in the first year or clustering on school attended in the first year*school advise*year) yield similar results. All regressions include lottery fixed effects. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. ***p<0.01 **p<0.05 *p<0.10. 26 With respect to socio-emotional functioning, independence at school has the only significant interaction effect. It indicates that students educated at school B score significantly lower on independence at school than their non-Montessori counterparts while no such effect is found for students participating in school admission lotteries for school A. Since 12 out of 14 interaction effects are insignificant, the general conclusion is that the effects of Montessori education are similar for school A and school B. Next to differences between schools, there might be differential effects of Montessori education for different groups of students. For example, it could be that girls do better at Montessori schools, since girls are generally more able to work independently than boys. For the same reason, it might also be that students in the higher school tracks do better in Montessori education than students in the lower school tracks. Tables C2 and C3 of Appendix C show additional differential effects of Montessori education for boys and girls and for different school tracks. Regarding gender, Table C2 shows that none of the interaction effects is significant at the 5% level. Therefore, the null-hypothesis of equal effects of Montessori education for boys and girls cannot be rejected, which implies that the effects of Montessori education are similar for boys and girls. For the different school tracks, some of the interaction effects are significant. The effects indicate that Montessori education has a more positive impact on prevocational IV students than on students at the academic tracks. Compared to senior general secondary/pre-university students, pre-vocational IV students more often obtain their degree on time on the advised track or higher. The main effects on this variable are not significant. Further, the positive effect of Montessori education on well-being at school seems to be driven by pre-vocational IV students. 27 6.4 Difference between lottery participants and students with priority As mentioned in Section 4.1, not every student attending the Montessori schools participated in a lottery. Both schools yield priority to students having brothers or sisters at the school and to children of employees. Moreover, school B conducts lotteries based on students’ residence, while school A yields priority to students from Montessori primary schools. It might be that students who got placed with priority are different from lottery participating students. Table 7 shows a-priori differences between students who participated in the lottery and students who got placed with priority. The table shows that on most background characteristics, there are no differences between lottery participators and students with priority. The differences that are found indicate that compared to students with priority, students who participate in the lottery are less often boy and more often have a non-western migrant background. More important than background characteristics is the fact that school A yields priority to students from Montessori primary schools. As the majority of the lottery participating students subscribed for school A, the effects of Montessori education are mainly identified on a sample of students who did not attend Montessori primary schools. It is possible that the effects of Montessori secondary education are more positive for students who were educated in Montessori primary schools: students from Montessori primary schools will have more experience with independent working and choice in their school tasks. Table 8 reports differences in the main outcomes between Montessori students who won the lottery and students who were placed with priority. It focusses on winning students as students who lost the lottery are generally educated at regular education schools. Although the results are descriptive, the results in the table indicate that the outcomes of students with priority are not significantly different 28 Table 7: Differences in background characteristics between students participating on the lottery and students being placed with priority Dependent variable Citoscore Pre-voc. IV primary school advice Pre-voc. IV/senior gen. primary school advice Senior gen. primary school advice Senior gen./pre-uni. primary school advice Pre-uni. primary school advice Boy Dutch Non-western migrant Western migrant Age at application Disadvantaged neighborhood Mean (SD) for Balancing test students with priority lottery participation (1) (2) (3) 541.0 0.128 0.497 (5.421) (0.189) 0.107 0.012 (0.310) (0.012) 0.082 -0.019 (0.274) (0.012) 0.248 0.031 (0.432) (0.021) 0.280 0.011 (0.449) (0.020) 0.283 -0.034 (0.451) (0.020) 0.408 -0.081 (0.492) (0.024) 0.744 -0.029 (0.437) (0.022) 0.132 0.039 (0.338) (0.018) 0.125 -0.010 (0.331) (0.016) 12.086 -0.003 (0.439) (0.021) 0.501 0.018 (0.500) (0.023) P-values 0.328 0.094 0.127 0.607 0.082 0.001 0.184 0.031 0.502 0.900 0.432 Number of students 1322 2122 Note: Column (1) displays the means and standard deviations for students who got placed with priority. Columns (2) and (3) report separate regression coefficients and the p-values of the dependent variables indicated in each row on an indicator variable equalling 0 if the student was placed with priority and equalling 1 if the student participated in the lottery. The regressions are on the full sample, including all students who participated in the lottery and all students who got placed with priority. All regressions include lottery fixed effects. Robust standard errors are reported in parentheses. 29 from the outcomes of students who won a school admission lottery.19 As the priority rules for school A are likely to create a larger difference between priority and non-priority students, it is interesting to note that the results are similar when only focussing on students at school A (results not shown). This implies that Montessori students who were educated at regular primary schools achieve similar to their secondary education peers who were educated at Montessori primary schools. Since students from Montessori primary schools do not get priority in school B, further support on this observation is provided by Table C1. This table shows that the effects of Montessori education are similar for students in school A and school B. Overall, the results in this subsection indicate that it is unlikely that the results on the effects of Montessori secondary education are driven by the priority rules. 7 Conclusions In this study, the causal effects of Montessori education are investigated by exploiting school admission lotteries for two Montessori secondary schools. It is found that Montessori education has little impact on students’ academic achievement. Montessori students obtain their secondary school degree without delay at the same rate and with similar grades as non-Montessori students, although the route towards the exams is somewhat different. After four years of secondary education, Montessori students more often attend a lower school track while non-Montessori students are more often grade-retained. On most socio-emotional outcomes, there is no significant effect of Montessori education. Most notably, there are no positive effects of Montessori education on independence and motivation, although these are the main characteristics Montessori education claims to foster. Regarding independence out of school, Montessori 19 The results for the other outcome measures (on grades and problematic behavior) are insignificant as well. 30 Table 8: Differences in outcomes between students winning the lottery and students being placed with priority Outcome (1) Achievement In 4th grade of advised track or better after four years -0.050* (0.029) In 4th grade of advised track or better after four years (cond. on being in 4th grade) -0.029 (0.024) Obtained degree on time on advised track or higher -0.031 (0.039) Obtained degree on advised track or higher with at most one year delay -0.027 (0.034) Enrolled in science program 0.020 (0.028) Parental support 0.263 (0.257) Participated in commercial school support 0.064 (0.054) Socio-emotional functioning Independence at school -0.992 (0.856) Independence out of school -0.195 (0.855) Motivation towards learning tasks -0.531 (0.474) Well-being at school -0.123 (0.704) Interest in society -0.164 (0.441) Note: Each coefficient reports an OLS regression regressing the dependent variable in that row on an indicator equalling 1 if the student participated in the lottery. The regressions include the students who got placed with priority and the students who won the lottery for the Montessori school. The number of students varies from 807 to 1381 for the academic outcomes and from 401 to 403 on the variables measured by the questionnaire. Robust standard errors are reported in parentheses. All regressions include lottery fixed effects and student controls. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. Exams are graded on a 10-point scale, where grades above 5.5 are considered a pass. ***p<0.01 **p<0.05 *p<0.10. 31 students even score worse than their non-Montessori counterparts. As the results in this study are based on two Montessori schools, one might worry that the results cannot be generalized to other Montessori secondary schools. It is reassuring that both schools in this study yield similar outcomes. Still, the results are based on Dutch Montessori secondary schools that conduct school admission lotteries. Since the demand for these schools is high, it might be that these are especially high quality Montessori programs. In that case, the results can be interpreted as the effects of Montessori education in high quality Montessori programs. Montessori secondary education programs are less standardized than Montessori primary education programs, and the results of this study should not be generalized to the small number of rather radical Montessori boarding schools with ’Erdkinder’ programs. The schools considered in this study are, however, part of a long tradition of Montessori secondary education. Montessori secondary schools in other countries are based on the same philosophy and have similar approaches with respect to choice of activities and field projects. Another issue is that the results are based on students applying for Montessori schools. It is likely that this is a selective sample of students, as students at Montessori schools are generally of higher socioeconomic status and less often from ethnic minorities (Dohrmann et al., 2007; Rathunde and Csikszentmihalyi, 2005a). It is possible that the effects of Montessori education are different for less advantaged groups of students. Further, the results for one of the schools are based on students who did not attend Montessori primary schools. From a critic’s point of view, you could argue that if Montessori education has a negative impact, the effects might be especially negative for students who did not attend Montessori primary schools. These students are unfamiliar with the Montessori education system and are probably not used to a large amount of choice and independent working. The outcomes of these students, however, turn out to be similar to the outcomes of the Montessori 32 secondary education students who were educated at a Montessori primary school. Despite its limitations, this is the first study to investigate the causal effects of Montessori secondary education. Overall, the results of this study indicate that Montessori education provides an alternative route to similar student outcomes. This is important information for parents, students and policy makers, as it does not support popular claims on the effects of Montessori education. Critics are not supported in their concern that Montessori education has a negative impact on student achievement. For advocates, on the other hand, there is no evidence that Montessori education has a positive effect on independence and motivation. 33 References Anderson, R. A., Worthington, L., Anderson, W. T., and Jennings, G. (1994). The development of an autonomy scale. Contemporary Family Therapy, 16(4):329– 345. Association Montessori Internationale (2013). Montessori movement. Retrieved 12-08-2013 from http://www.montessori-ami.org/. Beckert, T. E. (2007). Cognitive autonomy and self-evaluation in adolescence: A conceptual investigation and instrument development. North American Journal of Psychology, 9(3):579–594. Buser, T., Niederle, M., and Oosterbeek, H. (2014). Gender, competitiveness and career choices*. The Quarterly Journal of Economics. Calff, J. (1980). Van pionier tot mammoet: Het Amsterdams Montessori Lyceum 1930-1980. Amsterdam: Stichting Montessori 50. Chattin-McNichols, J. (1992). The Montessori controversy. New York: Delmar. Dohrmann, K. R., Nishida, T. K., Gartner, A., Lipsky, D. K., and Grimm, K. J. (2007). High school outcomes for students in a public Montessori program. Journal of Research in Childhood Education, 22(2):205–217. Karnes, M. B., Shwedel, A. M., and Williams, M. B. (1983). A comparison of five approaches for educating young children from low-income homes. In Consortium for Longitudinal Studies, editor, As the twig is bent: Lasting effects of preschool programs, pages 133–170. Hillsdale: Lawrence Erlbaum Associates. Karnes, M. B., Teska, J. A., and Hodgins, A. S. (1970). The effects of four programs of classroom intervention on the intellectual and language development of 4-yearold disadvantaged children. American Journal of Orthopsychiatry, 40(1):58–76. 34 Leuven, E., Oosterbeek, H., and Klaauw, B. (2010). The effect of financial rewards on students’ achievement: Evidence from a randomized experiment. Journal of the European Economic Association, 8(6):1243–1265. Lillard, A. and Else-Quest, N. (2006). The early years: Evaluating Montessori education. Science, 313(5795):1893–1894. Lillard, A. S. (2005). Montessori: The science behind the genius. New York: Oxford University Press. Lillard, A. S. (2013). Playful learning and Montessori education. American Journal of Play, 5(2):157–186. Miller, L. B. and Bizzell, R. P. (1984). Long-term effects of four preschool programs: Ninth- and tenth-grade results. Child Development, 55(4):1570–1587. Miller, L. B., Dyer, J. L., Stevenson, H., and White, S. H. (1975). Four preschool programs: Their dimensions and effects. Monographs of the Society for Research in Child Development, 40(5-6):1–170. Montessori, M. (1972). The discovery of the child. New York: Ballantine Books. Montessori, M. (1973). From childhood to adolescence. Including "Erdkinder" and the functions of the university. Second Edition. New York: Shocken Books. Montessori, M. (1989a). The child, society and the world. Unpublished speeches and writings. The Clio Montessori series volume 7. Oxford: Clio Press. Montessori, M. (1989b). What you should know about your child. Based on lectures delivered by Maria Montessori. The Clio Montessori series volume 4. Oxford: Clio Press. Mooney, C. G. (2000). Theories of childhood: An introduction to Dewey, Montessori, Erikson, Piaget, and Vygotsky. St. Paul: Redleaf Press. 35 Rathunde, K. and Csikszentmihalyi, M. (2005a). Middle school students’ motivation and quality of experience: A comparison of Montessori and traditional school environments. American Journal of Education, 111(3):341–371. Rathunde, K. and Csikszentmihalyi, M. (2005b). The social context of middle school: Teachers, friends, and activities in Montessori and traditional school environments. The elementary school journal, 106(1):59–79. Rubinstein, M. (2008). Karakteristieken van scholen voor voortgezet Montessorionderwijs [Characteristics of schools for secondary Montessori education]. Apeldoorn: Garant. Seldin, T. and Epstein, P. (2003). The Montessori Way. Sarasota: The Montessori Foundation Press. Smits, J.A.E. & Vorst, H. (2008). Schoolvragenlijst: Herziene uitgave 2008, Voortgezet Onderwijs [School questionnaire: revised edition 2008, secondary education]. Amsterdam: Pearson. Søgaard, A. J., Selmer, R., Bjertness, E., and Thelle, D. (2004). The Oslo Health Study: The impact of self-selection in a large, population-based survey. International journal for equity in health, 3(1):3. Van Loon, A. J. M., Tijhuis, M., Picavet, H. S. J., Surtees, P. G., and Ormel, J. (2003). Survey non-response in the Netherlands: Effects on prevalence estimates and associations. Annals of Epidemiology, 13(2):105 – 110. Vorst, H. C. M. (1993). Leren en studeren: Vragenboekje voortgezet onderwijs en middelbaar beroepsonderwijs [Learning and studying: Questions for secondary and vocational education]. Nijmegen: Berkhout B.V. Whitescarver, K. and Cossentino, J. (2008). Montessori and the mainstream: A 36 century of reform on the margins. The Teachers College Record, 110(12):2571– 2600. 37 Appendix A Details on questionnaire sampling and response In this appendix, the sampling procedure for the questionnaire and the questionnaire response are described. Because of the large difference in the number of lottery losing students, a different sampling procedure was used for school A and school B. Table A1 reports the number of lottery winners, losers and priority students selected for the questionnaire in each year, school and school track. Table A1: Number of students selected for the questionnaire No questionnaire School year Lottery winner Questionnaire Priority Lottery winner Lottery loser Priority Total School A: Senior general secondary/pre-university education 2003/2004 114 50 46 55 265 2004/2005 184 16 59 19 278 2005/2006 141 47 9 55 252 2009/2010 117 37 42 46 242 2010/2011 145 29 17 33 224 Total 701 179 173 208 1261 School A: Pre-vocational IV education 2003/2004 34 17 45 19 115 2004/2005 47 10 30 13 100 2007/2008 52 10 29 10 101 2008/2009 9 29 3 33 74 2009/2010 36 5 10 7 58 Total 178 71 117 82 448 School B: Senior general secondary/pre-university education 2008/2009 28 60 24 15 127 2009/2010 77 16 6 4 103 2010/2011 25 43 13 9 90 2011/2012 50 34 6 3 93 Total 180 153 49 31 413 Grand Total 180 1032 299 321 290 2122 Note: This table reports the number of students selected for the questionnaire by school, school track, year and lottery status. For school A, all 540 students participating in the school admission lotteries 38 were selected to receive the questionnaire. Besides that, a stratified random sample of 290 priority students was drawn. This sample is stratified by school year and school track, weighted on the number of lottery winners plus 10%. Since the number of students losing the lottery is much lower in school B, the questionnaire sample at school B was restricted to lottery participating students. Because the number of winning students is much larger than the number of losing students, a weighted random sample of winning students was used. This sample is weighted on the yearly number of lottery losers plus 40%, with a minimum of 6 students. This way, a total of 80 students applying to school B were selected for the questionnaire, bringing the total number of students selected for the questionnaire to 910.20 To protect the privacy of the students, the invitations for the questionnaire were sent out via the Dutch educational administration. Four different letters were sent: for Montessori and non-Montessori students and for students above and under 16. The last distinction was made because parental permission is required for research participation by students under 16 years of age. 505 students were sent the letter for Montessori students of 16 years and older, 98 students got the letter for Montessori students under 16. 268 students got the letter for lottery losing students of 16 years and older, 31 students got the letter for lottery losing students under 16. The online questionnaire was filled out by 609 students (67%), including 415 lottery participators. As described in Section 4.3, there was selective nonresponse to the questionnaire. Characteristics of responders and non responders are summarized in Table A2. 20 In fact, the original number of questionnaires requested to be sent out was 915 instead of 910. 5 students were dropped from the sample later on because of missing values on ethnicity and primary school advice. It turned out that 13 students did not have accurate address information, so 902 letters were sent. Which students have inaccurate address information is unknown. 39 Table A2: Selectivity in nonresponse to the questionnaire Non-responders Responders P-value School B 0.053 0.105 0.009 Admission for year 2003/2004 0.359 0.204 0.000 Admission for year 2004/2005 0.216 0.135 0.002 Admission for year 2005/2006 0.116 0.125 0.712 Admission for year 2007/2008 0.057 0.053 0.805 Admission for year 2008/2009 0.083 0.130 0.037 Admission for year 2009/2010 0.093 0.212 0.000 Admission for year 2010/2011 0.076 0.128 0.020 Admission for year 2011/2012 0.000 0.015 0.034 Lottery participant 0.681 0.681 0.991 Lottery loser 0.405 0.327 0.020 Citoscore 540.2 540.8 0.115 Pre-voc. IV primary school advice 0.263 0.130 0.000 Pre-voc. IV/senior gen. primary school advice 0.100 0.113 0.535 Senior gen. primary school advice 0.249 0.227 0.450 Senior gen./pre-uni. primary school advice 0.223 0.271 0.116 Pre-uni. primary school advice 0.166 0.259 0.002 Boy 0.392 0.322 0.036 Dutch 0.651 0.723 0.027 Non-western migrant 0.219 0.164 0.043 Western migrant 0.130 0.113 0.476 Age at application 12.166 12.071 0.002 Disadvantaged neighborhood 0.535 0.498 0.290 Number of students 301 609 Note: The first two columns report the mean of questionnaire non-responders and questionnaire responders for the characteristic mentioned in that row. The p-values refer to two-group mean comparison t-tests comparing students who did and did not respond on the questionnaire. 40 Appendix B First stage coefficients Tables B1 and B2 report the first stage coefficients belonging to the IV regressions in Tables 5 and 6. For all variables, the first stage coefficients are large (between 0.826 and 0.858) and significant at the 1% level. The partial F-statistics are high as well, ranging from 492.78 to 1083.18, well above the rule of thumb of a minimum of 10. These results confirm that the lottery is indeed a relevant instrument for presence at the Montessori secondary school. 41 Table B1: First stages achievement Outcome (1) (2) 0.844*** 0.843*** (0.026) (0.026) 0.848*** 0.848*** (0.027) (0.028) 0.857*** 0.858*** (0.029) (0.031) 0.867*** 0.869*** (0.027) (0.030) 0.856*** 0.858*** (0.029) (0.031) 0.856*** 0.857*** (0.029) (0.031) 0.856*** 0.857*** (0.029) (0.031) 0.837*** 0.838*** (0.034) (0.038) 0.845*** 0.846*** (0.029) (0.030) Parental support 0.835*** 0.827*** (0.030) (0.031) Participated in commercial 0.834*** 0.826*** (0.030) (0.031) In 4th grade of advised track or better after four years In 4th grade of advised track or better after four years (cond. on being in 4th grade) Obtained degree on time on advised track or higher Obtained degree on advised track or higher with at most one year delay Obtained degree on time on advised track with at least a 7.5 on average Obtained degree on time on advised track with passing grade for Dutch Obtained degree on time on advised track with passing grade for English Obtained degree on time on advised track with passing grade for Math Enrolled in science program school support ! Controls Note: Each row reports two sets of first stages of IV regressions with winning the lottery as an IV for ever attending a Montessori school. The sample is restricted to students participating on the lottery. Robust standard errors are reported in parentheses. All regressions include lottery fixed effects. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. ***p<0.01 **p<0.05 *p<0.10. 42 Table B2: First stages socio-emotional functioning Outcome (1) (2) 0.835*** 0.827*** (0.030) (0.031) 0.835*** 0.827*** (0.030) (0.031) 0.835*** 0.827*** (0.030) (0.031) 0.835*** 0.827*** (0.030) (0.031) 0.835*** 0.827*** (0.030) (0.031) 0.834*** 0.826*** (0.030) (0.031) 0.834*** 0.826*** (0.030) (0.031) 0.834*** 0.826*** (0.030) (0.031) Ever got a fine or was suspected of a 0.834*** 0.826*** criminal offense by the police (0.030) (0.031) Independence at school Independence out of school Motivation towards learning tasks Well-being at school Interest in society Number of times being late, truant or sent out of class in the past 6 months Smoked cigarettes or drank alcohol in the past month Used drugs of any kind in the past 12 months ! Controls Note: Each row reports two sets of first stages of IV regressions with winning the lottery as an IV for ever attending a Montessori school. The sample is restricted to students participating on the lottery. Robust standard errors are reported in parentheses. All regressions include lottery fixed effects. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. ***p<0.01 **p<0.05 *p<0.10. 43 Appendix C Heterogeneous effects Table C1: Differential results for school A and school B Outcome Treatment Interaction Treatment effect school A effect effect school B (1) (2) (3) -0.027 0.036 0.010 (0.049) (0.136) (0.131) -0.076* -0.038 -0.114*** (0.041) (0.055) (0.044) -0.018 -0.082 -0.100 (0.052) (0.161) (0.155) 0.329 0.473 0.801 (0.343) (0.716) (0.642) 0.048 0.386*** 0.435*** (0.072) (0.147) (0.132) Achievement In 4th grade of advised track or better after four years In 4th grade of advised track or better after four years (cond. on being in 4th grade) Enrolled in science program Parental support Participated in commercial school support (Table continues on next page) 44 Table C1: (continued) Outcome Treatment Interaction Treatment effect school A effect effect school B (1) (2) (3) 0.879 -4.778** -3.899** (1.048) (2.165) (1.913) -1.623 -1.828 -3.451* (1.082) (2.186) (1.948) -0.094 -1.061 -1.155 (0.627) (1.220) (1.054) 1.827* -2.223 -0.396 (0.941) (1.757) (1.503) 0.763 -0.628 0.135 (0.557) (1.169) (1.058) -7.131 7.880 0.749 (5.008) (6.700) (4.655) 0.062 0.121 0.183 (0.060) (0.129) (0.116) 0.016 -0.012 0.004 (0.067) (0.099) (0.075) 0.025 -0.008 0.017 Socio-emotional functioning Independence at school Independence out of school Motivation towards learning tasks Well-being at school Interest in society Number of times being late, truant or sent out of class in the past 6 months Smoked cigarettes or drank alcohol in the past month Used drugs of any kind in the past 12 months Ever got a fine or was suspected of a criminal offense by the police (0.059) (0.097) (0.076) Note: Each row reports two IV regressions with winning the lottery as an IV for ever attending a Montessori school. Columns (1) and (2) display the treatment effect for school A and the interaction effect on school B. Column (3) is another IV regression describing the treatment effect for school B. The sample is restricted to students participating on the lottery. Differential effects on exam measures are not available because the students participating in admission lotteries for school B are too young to have done secondary school exams. Robust standard errors are reported in parentheses. All regressions include lottery fixed effects and controls. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. Exams are graded on a 10-point scale, where grades above 5.5 are considered a pass. ***p<0.01 **p<0.05 *p<0.10. 45 Table C2: Differential results for boys and girls Outcome Treatment Interaction Treatment effect girls effect effect boys (1) (2) (3) 0.016 -0.115 -0.099 (0.054) (0.083) (0.074) -0.058 -0.073 -0.131** (0.042) (0.064) (0.057) 0.010 -0.030 -0.020 (0.071) (0.118) (0.101) 0.017 -0.029 -0.012 (0.064) (0.110) (0.098) -0.061 0.095 0.034 (0.053) (0.095) (0.092) 0.478 -0.215 0.264 (0.366) (0.591) (0.500) 0.164** -0.153 0.011 (0.080) (0.125) (0.103) 0.383 -0.858 -0.476 (1.132) (1.884) (1.559) -1.148 -2.308 -3.456** (1.143) (2.031) (1.721) 0.404 -2.003* -1.599* (0.657) (1.083) (0.910) 1.815* -1.051 0.764 (1.044) (1.616) (1.271) 0.320 1.011 1.331 Achievement In 4th grade of advised track or better after four years In 4th grade of advised track or better after four years (cond. on being in 4th grade) Obtained degree on time on advised track or higher Obtained degree on advised track or higher with at most one year delay Enrolled in science program Parental support Participated in commercial school support Socio-emotional functioning Independence at school Independence out of school Motivation towards learning tasks Well-being at school Interest in society (0.593) (1.006) (0.866) Note: Each row reports two IV regressions with winning the lottery as an IV for ever attending a Montessori school. Columns (1) and (2) display the treatment effect for girls and the interaction effect on boys. Column (3) is another IV regression describing the treatment effect for boys. The sample is restricted to students participating on the lottery. Robust standard errors are reported in parentheses. All regressions include lottery fixed effects and controls. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. Exams are graded on a 10-point scale, where grades above 5.5 are considered a pass. ***p<0.01 **p<0.05 *p<0.10. 46 Table C3: Differential results for pre-vocational IV and academic students Outcome Treatment Interaction Treatment effect effect effect pre-voc. IV senior gen./ pre-uni. (1) (2) (3) 0.048 -0.103 -0.054 (0.084) (0.101) (0.057) -0.039 -0.060 -0.099** (0.073) (0.085) (0.043) 0.142 -0.248** -0.106 (0.092) (0.120) (0.079) 0.079 -0.124 -0.044 (0.091) (0.117) (0.073) 0.015 -0.054 -0.039 (0.131) (0.144) (0.057) 0.293 0.153 0.445 (0.684) (0.760) (0.339) -0.054 0.224 0.170** (0.131) (0.152) (0.076) -0.142 0.318 0.176 (1.856) (2.111) (1.067) -2.006 0.112 -1.894* (2.107) (2.381) (1.095) 0.599 -1.166 -0.567 (1.215) (1.366) (0.621) 4.324** -3.847* 0.477 (1.781) (1.996) (0.915) 1.070 -0.551 0.519 Achievement In 4th grade of advised track or better after four years In 4th grade of advised track or better after four years (cond. on being in 4th grade) Obtained degree on time on advised track or higher Obtained degree on advised track or higher with at most one year delay Enrolled in science program Parental support Participated in commercial school support Socio-emotional functioning Independence at school Independence out of school Motivation towards learning tasks Well-being at school Interest in society (0.967) (1.139) (0.596) Note: Each row reports two IV regressions with winning the lottery as an IV for ever attending a Montessori school. Columns (1) and (2) display the treatment effect for pre-vocational IV students and the interaction effect on senior general/pre-university students. Column (3) is another IV regression describing the treatment effect for senior general/pre-university students. The sample is restricted to students participating on the lottery. Robust standard errors are reported in parentheses. All regressions include lottery fixed effects and controls. Controls include gender, being a nonwestern or western migrant, student age, living in a disadvantaged neighborhood, cito testscore, making another test than cito, no information on primary school test-score and primary school advice. Exams are graded on a 10-point scale, where grades above 5.5 are considered a pass. ***p<0.01 **p<0.05 *p<0.10. 47 TIER WORKING PAPER SERIES TIER WP 14/22 © TIER 2014 ISBN 978-94-003-0091-0
© Copyright 2025 Paperzz