WEBVTT

1
00:00:00.001 --> 00:00:04.389
Carlo BARONE: hmm. okay, I think we can. we can start.

2
00:00:04.610 --> 00:00:12.690
Carlo BARONE: So I guess I should say good morning to everyone for those connected from

3
00:00:13.030 --> 00:00:19.850
Carlo BARONE:  Melbourne. and good evening. Good night for the

4
00:00:20.170 --> 00:00:27.280
Carlo BARONE: people in in Europe and in the Us. anyway. Welcome to everyone. Hi!

5
00:00:29.260 --> 00:00:43.120
Carlo BARONE: Welcome to everyone! To these R. C. 28 session on a child development and education and inequalities we have, I think,

6
00:00:43.400 --> 00:00:53.280
Carlo BARONE: 3 or 4 presentations I don't see Kathryn bark on zoom I don't know if she's

7
00:00:53.330 --> 00:00:57.580
Carlo BARONE: she is in the room. Okay? Great.

8
00:00:57.610 --> 00:01:18.039
Carlo BARONE: So that means we have We have 28 min per session with the 4 presentations. So I would suggest, your presentation should take no more than 1718 min, so that we have 10 min for the discussion. Ideally.

9
00:01:19.100 --> 00:01:23.540
it's okay that I a chair here.

10
00:01:23.630 --> 00:01:32.429
Carlo BARONE: sure, of course, together with you, but otherwise you cannot see the room. So it is a kind of difficult.

11
00:01:32.470 --> 00:01:48.710
Carlo BARONE: It's good. We'll do it together. Many thanks. yeah. Yelsey was supposed to co-chair from Melbourne, but he wrote to me a few days ago that he couldn't attend, indeed!

12
00:01:48.750 --> 00:01:52.410
Carlo BARONE: So I wouldn't wait any longer, and

13
00:01:53.010 --> 00:02:08.859
Carlo BARONE: and And then start with the with the first presentation from team why do some children perform better than others genetic and environmental differences in in in learning growth?

14
00:02:09.490 --> 00:02:11.629
Carlo BARONE: So, Kim, please go on.

15
00:02:16.090 --> 00:02:22.450
Kim Stienstra: Yes, thank you. If everything shows correctly. Yes.

16
00:02:22.510 --> 00:02:24.269
Kim Stienstra: okay, then I will.

17
00:02:24.520 --> 00:02:42.059
Kim Stienstra: and I will start lower one I will share results of the last chapter of my dissertation which is on how Inequality develops during primary education, where I look into the genetic and environmental sources underlying dispersion and learning.

18
00:02:44.310 --> 00:02:54.949
Kim Stienstra: So large inequalities in educational performance can contribute to the diversions in life outcomes, including educational attainment income and several social and health outcomes

19
00:02:55.020 --> 00:03:06.910
Kim Stienstra: and important questions for public policy are therefore, why on children perform better than others, and how inequalities can be reduced well, prior studies often, study this

20
00:03:06.970 --> 00:03:26.939
Kim Stienstra: by looking at education performance at a single time points, or by investigating learning growth of the time. And often this is done by investigating this along the line of a script of background characteristics such as the socioeconomic backgrounds gender and different races and ethnicities.

21
00:03:27.190 --> 00:03:30.570
Kim Stienstra: and this is also be labeled as inequality of opportunity.

22
00:03:30.730 --> 00:03:43.949
Kim Stienstra: And though although this is important and insightful the emphasis on inequality of educational opportunity comes also at the expense of understanding the much broader question of how schooling affects.

23
00:03:43.960 --> 00:03:53.110
Kim Stienstra: The total variance of educational performance. This is another type of inequality which has been labeled as inequality, as dispersion.

24
00:03:53.720 --> 00:04:09.220
Kim Stienstra: well focusing on Inequality of this person seems particularly relevant, giving that described Chris directive characteristics, such as ses often explain only a small proportion of the variants and educational performance and learning codes

25
00:04:09.510 --> 00:04:15.050
Kim Stienstra: so many of the differences. This remain unexplained by these characteristics.

26
00:04:18.620 --> 00:04:31.900
Kim Stienstra: when children start schooling. They already differ considerably in their skill levels. And these differences reflect the different environments these children originate from, but also genetic differences between children.

27
00:04:32.430 --> 00:04:59.779
Kim Stienstra:  and I will just focus on the dispersion. So the total differences in performance and the underlying sources. I, where I rely on data on twins to distinguish 3 different sources of differences, namely, genetic differences, shared environmental differences which reflect all aspects that makes twins in the same same family more alike, such as the family they originate from, and the school they attend.

28
00:05:00.190 --> 00:05:08.469
and Lastly, noon shared a fine mental differences which include sources that makes twins more dissimilar from each other.

29
00:05:09.710 --> 00:05:30.629
Kim Stienstra: well, there are different normative standpoint. To what extent these different sources of this person are seeing as problematic. So in general dispersion. is often not seen as problematic, because a certain degree of variation and skills levels is needed because the labor market demands differently skilled workers.

30
00:05:31.390 --> 00:05:39.700
Kim Stienstra: but because this person can originate from different sources, and these sources can be seen as more or less problematic.

31
00:05:39.870 --> 00:05:48.860
Kim Stienstra: Also, some sources are more likely to be reduced or in time to to reduce during schooling than others.

32
00:05:49.490 --> 00:06:06.140
Kim Stienstra: So in general, these shared environmental differences, reflecting differences in family backgrounds. Amongst others, are seen as a problematic source or an unfair source of differences. But for genetic differences there's way less consensus.

33
00:06:06.380 --> 00:06:14.669
Kim Stienstra: So on the one hand, one could say that someone has no control on the day or on their genetic predisposition.

34
00:06:14.690 --> 00:06:19.129
Kim Stienstra: So just as their social background is, can be seen as an ascribed. Vector

35
00:06:19.700 --> 00:06:26.090
Kim Stienstra: and one might then think that this is a source of differences that you'll be compensated during schooling.

36
00:06:26.240 --> 00:06:40.160
Kim Stienstra: But, on the other hand, genetic differences also reflect differences in educational potential, which is something that one might argue is something that. it's allowed to influence once educational performance.

37
00:06:41.620 --> 00:06:56.989
Kim Stienstra: so in this study, I separate these different sources of differences and investigate to what extent they are reproduced, exacerbated, or compensated over the primary school career.

38
00:06:59.910 --> 00:07:07.830
Kim Stienstra: I investigate this, using data on of the nature in Netherlands cohort study and education.

39
00:07:08.120 --> 00:07:22.859
Kim Stienstra: in the Netherlands people's take a standardized achievement test during primary school. and as part of the average cohort study schools were asked to share these test results with a statistics. Netherlands.

40
00:07:22.980 --> 00:07:31.269
Kim Stienstra: as I have access to the reading and mathematics performance of many pupils throughout, create one till 5

41
00:07:31.800 --> 00:07:43.499
Kim Stienstra: where they taken the test twice a year, and that means I can rely on almost 10 measurements points of mathematics and reading achievements.

42
00:07:43.650 --> 00:08:03.149
Kim Stienstra: And this is the expenses by this for these types of study, that it minimizes a bias due to your question towards the name. And also it's it's that's an advantage of these test. Scores are that they are skilled based on item response theory. So they are comparable between school grades and different measurements moments.

43
00:08:04.300 --> 00:08:15.740
Kim Stienstra: So since these data are linked to the administered data, I've also information on many demographic backgrounds information and the school they these children attend

44
00:08:16.280 --> 00:08:32.570
Kim Stienstra: So I use the demographic information to identify to in Paris, and use the education and income of the parents to construct and family ses and aggregate it is to the school level to measure school as a ses.

45
00:08:34.110 --> 00:09:00.460
Kim Stienstra: Lastly, also, information on as the school environment is available. I link to data to many indicators of the dis inspective inspector of education. so I can measure the school quality which reflects different dimensions. such as, the range of educational activities that schools offer the learning, climate and social climate.

46
00:09:00.980 --> 00:09:03.549
Kim Stienstra: And I sophisticated to what extent?

47
00:09:03.570 --> 00:09:11.250
Kim Stienstra: these these school differences actually, can explain part of the of the differences.

48
00:09:12.830 --> 00:09:27.270
Kim Stienstra: So usually a twin design compares identical twins with fraternal twins. these 2 types differ in their genetic relatedness while they are assumed to share the shared refinement to the same extent.

49
00:09:27.340 --> 00:09:43.660
Kim Stienstra: but in registered data the psychos of the if twins is unknown. So therefore, I rely on the comparison of same sex and opposite extrins opposite sex. Twins are always fraternal. so their genetic relatedness is 50%.

50
00:09:44.050 --> 00:09:48.610
Kim Stienstra: Well, same-sex twins are a mixture of identical and fraternal twins.

51
00:09:48.790 --> 00:10:01.900
Kim Stienstra: And often it's it is assumed that this mixture is 50 50. So their average genetic relatedness can also be assumed to be the efforts of identical. And and so it's 55%

52
00:10:02.290 --> 00:10:11.510
Kim Stienstra: As for business takes. I also ferry this number to see how sensitive to results are to this impeded genetic relatedness.

53
00:10:11.900 --> 00:10:15.589
Kim Stienstra: but the results are very, very similar.

54
00:10:18.590 --> 00:10:36.419
Kim Stienstra: so based on the comparison of the educational performance of same sex and opposite sex twins. it is possible to use structural equation modeling to decompose the variance and educational performance into 3 components, the a component with which captures genetic differences.

55
00:10:36.720 --> 00:10:49.420
Kim Stienstra: the c components capturing shutter environmental differences and e components which captures not only share my afternoon share to find mental differences so inferences that are individual, specific

56
00:10:49.620 --> 00:10:55.890
Kim Stienstra: such as peer influences, for example. And this components also includes measurements, error

57
00:10:56.590 --> 00:11:03.499
Kim Stienstra: I applied this to in design to growth model, and this model consists of 2 late inspectors

58
00:11:03.520 --> 00:11:13.350
Kim Stienstra: first intercepts which estimates the average initial education performance at a start of formal education. So in grade one

59
00:11:14.400 --> 00:11:20.179
Kim Stienstra: and second, the slope which, estimates estimates the average learning growth.

60
00:11:20.780 --> 00:11:31.950
Kim Stienstra: and since the spacing between a different test is to be different this is reflected in the a different spacing between them. underlying pads.

61
00:11:32.850 --> 00:11:51.980
Kim Stienstra: the covariance between the intercept and slope shows whether a high initial performance have a stronger learning code or a recur learning route. And this reflects whether the initial dispersion. So the initial variance increase or decrease over time.

62
00:11:53.100 --> 00:12:01.420
Kim Stienstra: applying this model to the twin data allows for decomposing these variances of the intercept and slope.

63
00:12:01.710 --> 00:12:06.749
Kim Stienstra: and that leads to the biometric latent group model. So here we have.

64
00:12:06.870 --> 00:12:13.139
Kim Stienstra: and the estimated intercept and slope and 2 sets of ace components

65
00:12:13.580 --> 00:12:19.409
Kim Stienstra: the first ace components. that's a difference that intercepts

66
00:12:19.460 --> 00:12:24.769
Kim Stienstra: reflect the extent to which differences initial performance

67
00:12:24.780 --> 00:12:35.019
Kim Stienstra: can be attributed to genetic and environmental differences. Then we have the cross paths that show to what extent the same inferences also affect learning codes.

68
00:12:35.410 --> 00:12:46.919
Kim Stienstra: And lastly, there is a second set of A/C E farings components that reflects a unique genetic environmental inferences that only affects learning clouds.

69
00:12:46.940 --> 00:12:50.969
Kim Stienstra: but I know it associated with the initial performance levels.

70
00:12:54.970 --> 00:13:06.069
Kim Stienstra: So if we first look at the initial performance, so here's the estimated performance at the beginning of formal education in grid one

71
00:13:06.210 --> 00:13:17.630
Kim Stienstra: And what we see here is that most of the differences are related to a genetic differences for both reading and mathematics. I've been more so for reading and mathematics.

72
00:13:18.070 --> 00:13:34.430
Kim Stienstra: and it's important to be aware that these genetic moments does not reflect genetic determinism. But actually genetic inferences that that are expressed in correlation and interaction with the environment

73
00:13:35.370 --> 00:13:45.470
Kim Stienstra: and the shared and on share requirements explain the much. A smaller part of the differences in performance and our decre importance.

74
00:13:47.220 --> 00:13:52.460
Kim Stienstra: So next I investigate how these differences develop over time.

75
00:13:52.480 --> 00:14:06.150
Kim Stienstra: So if you first look at the differences in reading in total, the variance or in inequality as dispersion decreases during the primary school career.

76
00:14:06.330 --> 00:14:13.670
Kim Stienstra: And if you look at the separate components. We see that this decrease is attributable to a degrees in genetic variance.

77
00:14:13.800 --> 00:14:20.489
Kim Stienstra: Well, the change in the shared and on shared environmental variance is not statistically significant.

78
00:14:20.820 --> 00:14:33.399
Kim Stienstra: So children differ in a reading performance when they enter school largely because of genetic differences. And these genetic differences are just compensated over the primary school career.

79
00:14:34.730 --> 00:14:54.080
Kim Stienstra: for mathematics. it looks a bit different. So we can see that in total, the initial differences decrease over primary school career. But here this decrease is not due to a decrease in genetic differences, but in due to a decrease in shared and we'll share the environmental differences.

80
00:14:54.550 --> 00:14:59.299
Kim Stienstra: because here the change in genetic variance is not statistically significant.

81
00:15:02.620 --> 00:15:11.869
Kim Stienstra: But next to what happens to these initial differences during primary school, also, new sources of dispersion are coming into play.

82
00:15:12.260 --> 00:15:18.020
Kim Stienstra: and for boot reading and mathematics these are mainly a new genetic influences.

83
00:15:18.350 --> 00:15:25.790
Kim Stienstra: and when this is considered, the total dispersion in educational performance increases during primary school.

84
00:15:27.660 --> 00:15:46.290
Kim Stienstra: for the shared requirements which captures most others, families and school inferences. I we see that this is already relatively low, and it's our new, not new influence is coming into play. So this already suggests that the contribution of schools

85
00:15:46.330 --> 00:15:49.460
Kim Stienstra: is slightly small.

86
00:15:49.850 --> 00:15:56.690
Kim Stienstra: but I included, the measured Chris characteristics of the school and finance to further examine this.

87
00:15:58.310 --> 00:16:03.840
Kim Stienstra: and if I include the effects of school quality and school scs.

88
00:16:03.920 --> 00:16:19.420
Kim Stienstra: and also parental ses to take into account that the resources economic selection into school. I find that school quality is not statistically, significantly associated with neither initial performance nor learning growth.

89
00:16:19.650 --> 00:16:35.699
Kim Stienstra: and that's Children in higher scs, schools have an average higher initial performance and higher learning crowds. But these associations are rather small, and also don't account for much of the variance in educational performance and growth.

90
00:16:38.240 --> 00:16:52.729
Kim Stienstra: from a mathematic performance. I find higher initial performance in higher quality schools and high Scs schools. I've been no association with learning codes, and also here it barely accounts for at the variance in performance in growth.

91
00:16:56.240 --> 00:16:59.970
Kim Stienstra: so I have 3 conclusions. First.

92
00:17:00.010 --> 00:17:12.260
Kim Stienstra: the initial. Perform a differences in educational performance decrease of the primary school career The dispersion of the start of education is largely related to a genetic variance

93
00:17:12.280 --> 00:17:16.520
Kim Stienstra: and to a smaller degree also, to the shared environments

94
00:17:19.030 --> 00:17:35.820
Kim Stienstra: and a dispersion related to initial performance decreased during a primary school career for reading. This is because the initial genetic differences degrees and for mathematics. This is because the shared and one shared environmental differences decrease.

95
00:17:40.480 --> 00:17:51.570
Kim Stienstra: and So this shows that for mathematics, sources of inequality of opportunity are compensated during the school career.

96
00:17:51.610 --> 00:17:57.970
Kim Stienstra: well for reading there is more to pain in terms of decreasing social inequality.

97
00:18:00.890 --> 00:18:08.180
Kim Stienstra: Second, there are new sources of dispersion that come into play, and these are mainly new genetic influences.

98
00:18:09.550 --> 00:18:23.549
Kim Stienstra: and for mathematics also, to a smaller extent, loan sh to find mental influences. And when these new inferences are taking into account this lead to an increase in a total dispersion

99
00:18:23.720 --> 00:18:42.009
Kim Stienstra: these new genetic influences. can be understood from the process of genetic innovation. So, for example, the exposure to the New Learning activities and peer groups within school may lead to the activation and expression of genet genetic differences.

100
00:18:42.530 --> 00:18:49.479
Kim Stienstra: But although these innovation processes occur during the schooling period, it does not necessarily

101
00:18:49.530 --> 00:19:04.729
Kim Stienstra: reflects processes that are produced by the school environment. It could also be related to the family environment, or simply be an effect of biological maturation, and some more research is needed to provide more insight into this.

102
00:19:06.370 --> 00:19:17.330
Kim Stienstra: And, lastly, the role of schools in explaining this person is, likely limited. This is indicated by the relatively small contribution of the shared environment.

103
00:19:17.430 --> 00:19:25.669
Kim Stienstra: and in fact, it's a measured school course. Characteristics are only weekly associated with from a synchronous and experimental variance.

104
00:19:26.180 --> 00:19:38.760
Kim Stienstra: this could be related to a measurement limitations. So school factors should should be stable a systematically related to a journey code in order to explain variance

105
00:19:39.720 --> 00:19:51.700
Kim Stienstra: but it could also be related to to that. We do do not measure the aspects of the school environment that it may matter most so it could be that in a country such as the Netherlands

106
00:19:51.730 --> 00:20:01.199
Kim Stienstra: between school differences onward more limited. And it's more to be in school requirements, especially the teacher. That makes a difference.

107
00:20:01.750 --> 00:20:24.069
Kim Stienstra: this does not mean that schools do no play play a role in explaining this person as they are more complex ways including gene environment interactions and correlations. So that will be an interesting direction for future research to investigate, for example, to what extent these new genetic inference is coming to play

108
00:20:24.200 --> 00:20:34.169
Kim Stienstra: to that because of the exposure to the school of finance. So thank you for your attention, and I'm looking forward to for the questions and points for this question.

109
00:20:41.500 --> 00:20:57.610
Carl, do you want to do thank you this immediately or at the end. Do you want to collect questions from the floor? First, maybe. Yeah. Okay, so we we do. After each presentation. We do immediately. questions. Okay, I will. I will do that. Yeah.

110
00:20:57.770 --> 00:21:02.870
yeah, okay, that's fine. Kim. If you would unshare, I can see everyone.

111
00:21:03.080 --> 00:21:09.030
and I can see the people in the room. If there is someone in the room who has a question.

112
00:21:09.280 --> 00:21:14.380
because otherwise you can start with Julia with Julie, who has a question online.

113
00:21:14.390 --> 00:21:15.430
Julie PEREIRA: Yes.

114
00:21:15.730 --> 00:21:17.930
Julie PEREIRA: thank you so much for your presentation.

115
00:21:18.290 --> 00:21:20.220
Julie PEREIRA: Can you hear me?

116
00:21:20.310 --> 00:21:32.930
Julie PEREIRA: Yes, thank you so much for your presentation. I have many questions, but I will limit myself to let other people speak. I guess my first question is, I just read your paper in

117
00:21:32.960 --> 00:21:51.309
Julie PEREIRA: European Sociological Review in 2,021, and I was wondering whether the order of magnitude of what is the relative part of genetics versus environmental influence. I have the feeling that they vary quite a bit between those 2.

118
00:21:51.350 --> 00:22:14.670
Julie PEREIRA: research. I had in mind something like 70 for this presentation for genetics versus 40, only for your other paper. So I was wondering, whether you have any insight on how stable those numbers are, and how you explain. this difference. Thank you so much.

119
00:22:14.780 --> 00:22:37.160
Kim Stienstra: Yes, thank you for your question. so this relative contributions of genetic differences to an outcome. they are not fixed, but very much contact context dependence. So there are also different many different ways how this could ferry between countries, but also between different outcomes and generally.

120
00:22:37.160 --> 00:22:54.099
Kim Stienstra: there are also differences between whether you look at educational attainments and educational performance, for example, and often This relative contribution of jeans is a bit lower for educational attainments and educational performance. as for educational attainments, also more

121
00:22:54.110 --> 00:22:57.659
Kim Stienstra: strongly social influences play a role

122
00:22:57.750 --> 00:23:01.269
Kim Stienstra: so that explains one of the difference. Differences

123
00:23:01.690 --> 00:23:24.190
Kim Stienstra: but also country differences into for with respect to how open educational system is, or how big social inequalities are in society. they also affect the relative contributions of teams and environment. So it's certainly not to be expected that this is the same percentage across different studies and different countries.

124
00:23:26.870 --> 00:23:30.359
Okay, okay, other questions in the room.

125
00:23:32.610 --> 00:23:54.510
Carlo BARONE: So people are very silent. So if there are questions, I could you elaborate a bit more on the different patterns for math and and and reading why, they, the the trends. What is the substantive interpretation for the differences in the trends over time. So to say.

126
00:23:54.680 --> 00:24:17.030
Kim Stienstra: yeah. So what we see is that for reading, the decrease is much more produced by a decrease in genetic difference as well for mathematics. This is due to differences in environmental differences. and I interpreted this as like, often. you see, that reading performance is something that is really

127
00:24:17.030 --> 00:24:37.040
Kim Stienstra: produce and shape within the home environments. so there are already a lot of processes that are going on in parents stimulating their children and also nurturing their genetic pre predisposition to reach. Well, So when they start schooling they are already much more Differences also kept it within this genetic components

128
00:24:37.110 --> 00:24:47.709
Kim Stienstra: which then might be, compensating to compensated during school. So, for example, genetically, tenants of children who don't get

129
00:24:48.310 --> 00:24:59.710
Kim Stienstra:  by their parents not get so much simulation, and they may not express his genetic, but when they enter school they may experience this more

130
00:25:00.070 --> 00:25:06.619
Kim Stienstra: well for mathematics. This is typically your domain that is more learned within school and a bit less in families.

131
00:25:06.730 --> 00:25:11.459
Kim Stienstra: so maybe here, this place like this whole G in your finance

132
00:25:11.570 --> 00:25:23.389
Kim Stienstra: into play plays a bit less a role in the family of finance that operates more during schooling. so these are myself. As to why this I might differ between reading and mathematics.

133
00:25:24.080 --> 00:25:25.020
Carlo BARONE: Thank you.

134
00:25:27.600 --> 00:25:30.360
Yes, there is a question here in the room.

135
00:25:30.550 --> 00:25:31.220
So

136
00:25:34.760 --> 00:25:44.459
Hi! It might be a bit of our own third question, so let me know if it is, and don't worry. But I was wondering what is the not value. But what is the meaning of using

137
00:25:44.470 --> 00:25:50.399
the word genetic? When we already say that it has to do with the kind of interplay between genetic

138
00:25:50.580 --> 00:25:57.830
material, or whatever the correct word is, and the environment, and especially in light, of what you showed us in terms of the new.

139
00:25:58.160 --> 00:26:14.530
Let's call it a new inequality that develops over school that has a genetic component. But then, of course, it has to do with things that are happening over the life course of the child. And so how much of this? Maybe the genetic material is there, but is not allowed to express.

140
00:26:14.600 --> 00:26:22.750
as it could, due to the environment. So I was just wondering if you had any thought on the use of genetic in this contest, thank you.

141
00:26:23.090 --> 00:26:44.900
Kim Stienstra:  yeah, so 1 one part license to separated, separating this genetic components, which also captures this complex interplay between gene and environments from the shared a finance, the components that reflects only differences in the environment of which we

142
00:26:45.160 --> 00:26:58.629
Kim Stienstra: it's often yeah, there's more consensus that this is something that should no play a role. So if we teach that to stand out. We can focus more on this part and see how this this changes during the school career.

143
00:26:59.090 --> 00:27:01.920
Kim Stienstra: and I think also

144
00:27:02.490 --> 00:27:09.519
Kim Stienstra: the there. There is no more of a consensus on To what extent these genetic differences?

145
00:27:09.580 --> 00:27:15.480
Kim Stienstra: to prep play a role. So it's also, I think, a way to

146
00:27:16.140 --> 00:27:25.279
Kim Stienstra: erez agmoni focus a bit more on this, and actually have more of this conversation of okay, this increase in genetic inferences, these new genetic inferences that come into play. 101

147
00:27:25.560 --> 00:27:29.569
Kim Stienstra: This might be a good thing if if this reflects

148
00:27:29.640 --> 00:27:54.260
Kim Stienstra: the genetic potential of children or genetic tenants that are expressed during schooling. but they could also reflect other sources of genetic differences. they might be bit less at that if these get expressed during schooling. So it also invites all the research to dive a bit more into into this, like, what actually does this capture? and have a bit more of this

149
00:27:54.300 --> 00:27:58.660
Kim Stienstra: a conversation of to what extent is to be

150
00:27:59.750 --> 00:28:01.830
Kim Stienstra: increase during the school career.

151
00:28:04.850 --> 00:28:11.200
I'm sorry, I think, Carlo, if you agreed that we need to go to the second. presentation. There are

152
00:28:11.380 --> 00:28:13.379
2 more questions.

153
00:28:13.640 --> 00:28:20.589
 so I'm sorry to call it. But maybe you could ask your question, and and then we'll see

154
00:28:20.840 --> 00:28:24.130
you quickly. Ask your question, and then

155
00:28:24.940 --> 00:28:25.720
hmm!

156
00:28:26.410 --> 00:28:34.329
My name is Takashiasakura. I came from Yakam by working at the teacher training university

157
00:28:34.360 --> 00:28:43.640
in the front. So, according to your research, if the genetic factor take

158
00:28:43.670 --> 00:28:58.669
take a a greater role than environment, the Iraq asks you the what is the role of a teacher, or in school to decrease the disparity.

159
00:28:58.840 --> 00:29:00.830
but in new equality.

160
00:29:02.150 --> 00:29:18.749
Kim Stienstra: so came in half a minute.

161
00:29:18.870 --> 00:29:25.590
Kim Stienstra: especially large amounts of genetic differences, and that these hardly changed during the primary school career.

162
00:29:25.600 --> 00:29:35.559
Kim Stienstra: so maybe one should lower the expectations a bit. With respect to to what extent schools and teachers could have really in compensatory

163
00:29:35.770 --> 00:29:58.169
Kim Stienstra: as the large amounts of differences aren't just, we produce during schooling, and maybe these more complex gen requirements into place our ways through which teachers could still play a role. But it won't be the case I think, based on these results that they would largely compensate already existing inequalities that are there at the start of schooling.

164
00:30:00.550 --> 00:30:07.500
Ok, thank you very much. I think we go to the second presentation, Philip, I assume you do that or not.

165
00:30:10.500 --> 00:30:20.269
you are going to present. Yeah, yeah, okay, very good.

166
00:30:21.290 --> 00:30:24.799
Julie PEREIRA: So thank you. Can you, can everyone hear me? Okay.

167
00:30:25.370 --> 00:30:38.690
Julie PEREIRA: yeah, it's okay. Cool. So thank you very much for being here. I'm really happy to be presenting today. Co. Joint work with Philip, that is also online with us.

168
00:30:38.890 --> 00:30:55.040
Julie PEREIRA: so the title of the presentation is early in intensive music education, what it packs and for whom and we are presenting today the evidence from a quizzy experiment of of a program called of Zoom Dominic Code, which means inviting in my school.

169
00:30:55.440 --> 00:31:05.680
Julie PEREIRA: So I'm first going to frame the research questions within the literature that I'm going to present the data methods, results and discussion.

170
00:31:06.630 --> 00:31:08.380
Julie PEREIRA: So

171
00:31:09.110 --> 00:31:22.780
Julie PEREIRA: the existing literature is very much focused on what we call extrinsic effects of arts. Education through the cognitive transfer paradigm. So what that means is, usually we divide

172
00:31:23.330 --> 00:31:45.900
Julie PEREIRA: impacts of art education into 2 different types of impacts. First, one is in in pink here. So it's intrinsic effect. So it's everything that that is directly linked with ours, education, so it can be things like art, enjoyment, artistic skills, and knowledge, etc. And then, what people are usually more interested in

173
00:31:45.900 --> 00:31:59.449
Julie PEREIRA: or in extrinsic effects, and those could be cognitive so that can be schooling performance, math, reading, etc., or and a non cognitive effects, such as empathy, self, confidence, etc.

174
00:32:00.100 --> 00:32:02.340
Julie PEREIRA: And usually

175
00:32:02.480 --> 00:32:19.900
Julie PEREIRA: there is like every research that is linked to the face of art. Education often has an attempt to justify the need for arts education in a context of limited funding. So this pushes researchers and political

176
00:32:20.200 --> 00:32:28.260
Julie PEREIRA: policymakers to hold a great importance, a greater in importance to extrinsic effects compared to intrinsic events.

177
00:32:29.580 --> 00:32:56.909
Julie PEREIRA: also, they are very few sociological studies. this is from a systematic literature review I I conducted, and you can see that the main disciplines here are psychology and education sciences. And then to a lesser extent, neuroscience is economics and arts. So everything that is directly linked to the art discipline. But we're still lacking a a strong sociological perspective on these matters.

178
00:32:58.730 --> 00:33:09.269
Julie PEREIRA: What we know is that music education is indeed commonly associated with lots of alleged benefits. so those are improved. IQ scores

179
00:33:09.280 --> 00:33:14.509
Julie PEREIRA: special time for our reasoning, reading, numeracy, social skills, etc.

180
00:33:14.800 --> 00:33:20.859
Julie PEREIRA: But there's been reset met analysis and overall, and

181
00:33:21.030 --> 00:33:21.900
Julie PEREIRA: oh.

182
00:33:22.080 --> 00:33:24.729
Julie PEREIRA: conflicting evidence on whether

183
00:33:25.040 --> 00:33:27.130
Julie PEREIRA: the link is causal or not.

184
00:33:27.530 --> 00:33:40.530
Julie PEREIRA: E, even in the meta analysis it's really interesting, because some of them have the same corpus, but they still reach different conclusion as to whether music education is a coastal impact on those could need to

185
00:33:40.720 --> 00:33:42.030
Julie PEREIRA: good news of skills.

186
00:33:43.700 --> 00:33:55.069
Julie PEREIRA: So our research questions are twofold. First, one is, can music education positively affect cognitive and non-cgnitive skills of young, socially disadvantaged children.

187
00:33:55.130 --> 00:34:07.770
Julie PEREIRA: And then, if such effects are demonstrated, how do they vary along? Students, characteristics just gender H and socioeconomic background, that is, who would benefit most from such an interventions.

188
00:34:09.529 --> 00:34:10.740
Julie PEREIRA: Here we have

189
00:34:10.890 --> 00:34:36.209
Julie PEREIRA: 2 main caricatural theoretical frameworks to simplify. So first, we would have the social cognitive perspective, so that would be within the framework of cultural mobility. That's the module or also yeah, cultural compensation. And so this perspective is interested in the social conditions of redistribution

190
00:34:36.210 --> 00:34:56.109
Julie PEREIRA: through arts, education of cultural and community resources to population that are perceived as the most deprived with regard to school. and this perspective is implicitly dominant in most most evaluating works. And then we have the second perspective that's more critical.

191
00:34:56.170 --> 00:35:09.649
Julie PEREIRA: that relies on on authors such as Belgium and pestle that really warrants about the amb ambivalence of such pedagogical actions, because those rely on a lot of social and cultural architecture.

192
00:35:09.890 --> 00:35:10.830
Julie PEREIRA: So

193
00:35:11.080 --> 00:35:14.430
Julie PEREIRA: this perspective, the second perspective.

194
00:35:15.760 --> 00:35:27.230
Julie PEREIRA: things that maybe those programs can actually just reinforce the expected compliance with the norms of the legitimate culture rather than really change

195
00:35:27.330 --> 00:35:28.789
Julie PEREIRA: community resources.

196
00:35:29.890 --> 00:35:33.250
Julie PEREIRA: So the program we are evaluating is

197
00:35:33.260 --> 00:36:00.739
Julie PEREIRA: interesting because it's quite intensive and also very early in life. So it's a program that teaches for the violin to kids from 4 to 8 years old. So 2 years in kindergarten, and then 2 years in primary school. and so they have quite a intensive training, because they have so first once a week for 45 min. But then the the 3 following years. They have 3

198
00:36:00.830 --> 00:36:08.620
Julie PEREIRA: classes a week. So one in a large group and 2 in in smaller groups. So that's about 2 h of violin. a week.

199
00:36:08.950 --> 00:36:23.669
Julie PEREIRA: This program is founded by a private philanthropic foundation. It is mandatory, so there is no effect of self selection. It takes place during school hours, and it involves every student new school, which is also a great parameter of interest.

200
00:36:24.940 --> 00:36:42.010
Julie PEREIRA: I suggest that just the timeline of our evaluation. So here we are here at this dotted line. so we collected launches in that data. both ourselves. So that's the great, the sorry, the gray line

201
00:36:42.020 --> 00:36:50.080
Julie PEREIRA: and then we also rely on that data was connected at the national level. So standardized testing

202
00:36:50.220 --> 00:36:51.769
Julie PEREIRA: and in green here.

203
00:36:53.380 --> 00:36:59.420
Julie PEREIRA: So what we are trying to measure, we have 3 sets of of different outcomes that we're interested in.

204
00:36:59.470 --> 00:37:03.050
Julie PEREIRA: The first one is cognitive and academic skills.

205
00:37:03.260 --> 00:37:13.069
Julie PEREIRA: And as I said, we're both using adult testing that we did individually and age children. And then academic tests that are more standardized.

206
00:37:13.780 --> 00:37:25.900
Julie PEREIRA: We also have an interest in assessing the social, emotional skills of these children. So we're for this purpose. We're using a short version of the Big 5 for kids that is filled by the teachers.

207
00:37:27.170 --> 00:37:46.650
Julie PEREIRA: We then want to try to assess musical skills, even though that's not the main point of the program, according to the philanthropists, and the policymakers. But it will still be interesting to determine whether some progress can be linked to musical progress, or if if it's something else. So the closer link.

208
00:37:47.070 --> 00:37:58.499
Julie PEREIRA: And then we also have social demographics. to contextualize everything. So we collected that through family survey, and we got about 80%. answer rates

209
00:38:00.100 --> 00:38:11.429
Julie PEREIRA: additional to that, we are also conducting, conducting qualitative equality for a survey. To better understand how this program is at originate.

210
00:38:11.470 --> 00:38:21.649
Julie PEREIRA: It's a originally incidented to understand how it fits within this school curriculum. what is being done? Less because of the program

211
00:38:21.690 --> 00:38:32.149
Julie PEREIRA: poets? how we, the program reception is on the teacher's side and the family side, and then find you in our ipothesis, regard to what

212
00:38:32.170 --> 00:38:34.820
Julie PEREIRA: are the mechanisms that are at play here?

213
00:38:36.630 --> 00:39:01.630
Julie PEREIRA: So the we arrived to a first problem. It's that we could not randomize. So our test group and our control group are not perfectly similar. So here you can say, a set of measure about reach. The the difference between those 2 purpose significant. It's not huge, but it's significant. And so overall, our control group is a little bit more advantage.

214
00:39:01.630 --> 00:39:10.220
Julie PEREIRA: then our test proof. So they start being a little bit better, and there is no way we can rule out that they would have progress

215
00:39:10.240 --> 00:39:21.720
Julie PEREIRA: a slightly faster than our test group. and additionally, we see that. So the first time is is called It's this, a social position index that relies on the

216
00:39:22.020 --> 00:39:28.700
Julie PEREIRA: on the jobs of the both parents. And we we also see that the data we collected

217
00:39:28.740 --> 00:39:31.750
Julie PEREIRA: is mostly with regards to the

218
00:39:31.800 --> 00:39:42.490
Julie PEREIRA: or maybe above the average social level of their school, because the real average, as computed by the ministry is, is lower than what we measured. So

219
00:39:42.660 --> 00:39:45.149
Julie PEREIRA: we also have to take that into account.

220
00:39:45.780 --> 00:39:58.380
Julie PEREIRA: So to deal with that because we don't have a randomized control trial, we are, we explored, and every quasi experimental method that that there is. So

221
00:39:58.890 --> 00:40:18.430
Julie PEREIRA: we we did use differences in differences. but we have the 2 problem. First, some outcomes could not be measured, accented because children were super young. They were barely 4 years old, so it's very hard to measure some things, and also we have no knowledge on whether the pre treatment trends were parallel

222
00:40:19.430 --> 00:40:26.280
Julie PEREIRA: then the regression. Discontinuity method is not applicable here. So we turned to matching methods.

223
00:40:27.990 --> 00:40:56.029
Julie PEREIRA: among matching methods we picked entropy balancing, which is a desktop, pre pre-processing method that is, a used to achieve covariance in observational studies with binary treatments. So there are 2 steps in this method. First step you peak a set of relevant covariates that you believe can influence the outcome you're measuring, and you pick a set of moments. So that is mean variance here, etc.,

224
00:40:56.270 --> 00:40:58.379
Julie PEREIRA: for which you want the balance to be done.

225
00:40:58.950 --> 00:41:03.179
Julie PEREIRA: and it then so at this step you want to.

226
00:41:03.600 --> 00:41:29.889
Julie PEREIRA: because to have a trade off. Either you include every variable, and you have a perfect balance and everything. But then you will have more less of a sample size, because you will have more missing values. So there's this trade off to to to to do. And then, if there exists a set of weights that parents is the true groups regarding what you you picked then entropy. Balancing is going to find it.

227
00:41:29.890 --> 00:41:54.639
Julie PEREIRA: So it's a it's an interesting method, I think, because there is no need to iterate manually the methods contrary to other waiting met techniques such as propensity for waiting, for instance. And also you don't have to drop observations. And in our case we had about 2,000 kids to start with, with. So not a huge sample so we could not afford to lose too many kids.

228
00:41:55.710 --> 00:42:02.780
Julie PEREIRA: So we picked this entropy balancing scores that includes gender

229
00:42:02.860 --> 00:42:13.969
Julie PEREIRA: initial cognitive scores trimester of birth because they're super young. So the age in month is really important. And then, the social positioning index

230
00:42:15.380 --> 00:42:28.670
Julie PEREIRA: I here I will just display results for this set of ways. But it doesn't change drastically if I include additional variables here. Well, basically, what changes are the

231
00:42:28.830 --> 00:42:30.870
Julie PEREIRA: confidence intervals mostly.

232
00:42:31.550 --> 00:42:33.859
Julie PEREIRA: And then

233
00:42:33.880 --> 00:42:54.529
Julie PEREIRA: sorry. I'm using multi-level mixed effects models to account for the fact that our kids are clustered within schools. and so I have a viding variable that takes 0 one whether this students treated or not. And then I controlled, for I control for the results.

234
00:42:54.940 --> 00:43:02.400
Julie PEREIRA: on the same outcome before the program. If I have this data. and then you have the random part of the model.

235
00:43:03.780 --> 00:43:06.800
Julie PEREIRA: So now on to the results.

236
00:43:06.860 --> 00:43:14.540
Carlo BARONE:  3 did. And control students means students from different schools. Right?

237
00:43:14.760 --> 00:43:24.889
Julie PEREIRA: Yes. sorry. Yeah. The treatment is assigned at the school level. So if If one school is treated, then all the kids inside the school are treated

238
00:43:26.170 --> 00:43:36.599
Julie PEREIRA: that some kids do move between schools or they move out. So I. It's still at this, at the children level the years of validated.

239
00:43:37.320 --> 00:43:40.729
Julie PEREIRA: And so as for the results, we have

240
00:43:41.180 --> 00:43:57.690
Julie PEREIRA: in the middle of the second year of the program. So one year and a half into the program, we do find, a positive substantial positive effect on written skills and a quasi significance. It impacts on math.

241
00:43:59.510 --> 00:44:03.090
Julie PEREIRA: I will. I will talk about why later on.

242
00:44:03.820 --> 00:44:05.730
Julie PEREIRA: But we also find.

243
00:44:06.600 --> 00:44:12.869
Julie PEREIRA: not significant, but barely negative impact on words understanding

244
00:44:15.290 --> 00:44:18.690
Julie PEREIRA: anyways the cognitive tests we run.

245
00:44:19.340 --> 00:44:28.390
Julie PEREIRA: Kids are still way behind the to your theory distribution, because it's a disadvantage neighborhood. And so they have about

246
00:44:28.450 --> 00:44:36.110
Julie PEREIRA: one year. they're about one year late. as to what they they, they they could be

247
00:44:36.230 --> 00:44:55.699
Julie PEREIRA: both in terms of yeah. So you have the initial words understanding results on the left and the after a one year on, on up to 1.5 year on on the right. And you see that they they did improve because they they're older. But They did not catch up being late.

248
00:44:56.160 --> 00:45:18.269
Julie PEREIRA: So same for another test, and then for non commutative trades, it seems that teachers assess their children to be more contentious which is a a good predictor for cognitive skills. That's that's a rather good sign. But we have to be careful because it's still assessed by the teacher. So question to

249
00:45:18.330 --> 00:45:19.580
Julie PEREIRA: injectivity.

250
00:45:20.700 --> 00:45:25.890
Julie PEREIRA: So far we find no impact on musical listing skills, which is rather surprising.

251
00:45:25.970 --> 00:45:36.639
Julie PEREIRA: but we are we testing the very same test right now? because we think maybe some, some. Some of this has to do with the fact that

252
00:45:37.140 --> 00:45:46.539
Julie PEREIRA: They had to understand the concept of being the same or being different, and maybe they were not old enough to grasp that. So maybe you will find effects after 3 years.

253
00:45:47.600 --> 00:46:01.999
Julie PEREIRA: So for the whole population, we have positive effect on rich and skills. Maybe because fine electricity was was trained more intensively. maybe because they started reading simplified music sheets. And that train their reading skills.

254
00:46:02.110 --> 00:46:15.339
Julie PEREIRA: We have a positive effect on consciousness that could be a flexible effect or creepy because the straits was trained more intensively. But we have no effect of musical skills and a negative effect on word understanding.

255
00:46:16.160 --> 00:46:20.940
Julie PEREIRA: When we regard when we look at subpopulation. we have

256
00:46:21.150 --> 00:46:24.140
Julie PEREIRA: same cognitive impacts for boys and girls. So far.

257
00:46:24.550 --> 00:46:30.519
Julie PEREIRA: a slightly more positive impact for non cognitive

258
00:46:30.540 --> 00:46:39.249
Julie PEREIRA: trades which is in line with their literature. possibly because girls had a a ceiling effect like less less room to grow.

259
00:46:40.660 --> 00:46:51.140
Julie PEREIRA: And that's interesting. By a social positioning index we find that the effects are stronger at the bottom of the distribution. So the more disadvantaged the kids the

260
00:46:51.330 --> 00:46:55.720
Julie PEREIRA: the better program. So that was for kindergarten

261
00:46:56.280 --> 00:47:19.849
Julie PEREIRA: and right now we are in the process of analyzing the the results for K. One. So first first year primary school, and we were quite surprised to find that actually, we find this time a negative substantial negative impact for for math and for language. It's not significant, but it seems to be quite negative.

262
00:47:20.250 --> 00:47:39.609
Julie PEREIRA: and the qualitative part of what we do seems to imply that it's probably because the program takes so much time during school week that kids actually have less time devoted to math and and language. And so, even though music might be beneficial. It's not

263
00:47:39.690 --> 00:47:48.830
Julie PEREIRA: beneficial enough to to balance the fact that they are losing instruction time in in other fundamental subjects.

264
00:47:49.260 --> 00:47:52.010
Julie PEREIRA: So thank you very much for your intention.

265
00:48:01.220 --> 00:48:03.469
hey? Thank you all questions.

266
00:48:04.480 --> 00:48:06.710
Maybe I can start with a

267
00:48:06.940 --> 00:48:12.069
a clarification question, because at the end I was a bit overwhelmed with the results.

268
00:48:12.170 --> 00:48:16.329
And I I wonder, did she now say that the effects?

269
00:48:16.430 --> 00:48:22.199
It was stronger for higher social classes, children or lower social classes? Children?

270
00:48:22.440 --> 00:48:42.360
Julie PEREIRA: Okay? So it's 2 full. So in kindergarten, it's higher, positive effect for lower. ses kids. So we have a registered achieve impact of the program. But after in primary school the negative impact seems to be stronger for kids,

271
00:48:42.850 --> 00:48:48.850
Julie PEREIRA: for more disadvantage backgrounds. So it's in both directions.

272
00:48:50.140 --> 00:49:01.789
Okay, yeah, thank you. But that was preliminary, wasn't it? Okay? So maybe that change there is this piece coming to the microphone. There's question from the in the room.

273
00:49:05.320 --> 00:49:11.849
Hi. This is Marit Ayaka from University of Cologne. I have a question regarding the research design.

274
00:49:11.910 --> 00:49:21.510
As far as I understood, you did not have a proper randomization due to the foundation. Who is given whatever money instruments into the school?

275
00:49:21.780 --> 00:49:47.180
But I would like to know a little bit more about the schools, so did you have? students in schools that have experience with the program? Or are these new schools where the program has been established for the first time for these students, or sees kids in kindergarten? This is the first question, and the second question, which is, which is related to that is.

276
00:49:47.600 --> 00:50:11.619
how can you be sure that it actually is the music classes, or hearing and doing music? That makes a difference, but rather the development in the kindergarten and in the schools where you have may have a different school climate, the community experiences during the classes. So how do you deal with this differences at the school level. That may explain part of your results.

277
00:50:11.790 --> 00:50:20.450
Julie PEREIRA:  yeah, thank you for a question, so for the research design, we we work with about 60 schools.

278
00:50:20.770 --> 00:50:29.079
Julie PEREIRA: and among those 60 schools we have approximately 40 schools that are violent schools and 20 control schools.

279
00:50:29.460 --> 00:50:45.270
Julie PEREIRA: those control schools there were picked, and the criteria were that they had to be labeled as priority education in the French system, and that they had to be located in the same cities than the test schools.

280
00:50:45.500 --> 00:50:59.820
Julie PEREIRA: so in that sense they are comfortable in that sense, but they also have to be voluntary, to to like, to agree to participate in in the research. And that's why we couldn't hire more control of

281
00:51:00.060 --> 00:51:06.380
Julie PEREIRA: And inside the violin schools 5 of them had started the program already.

282
00:51:06.640 --> 00:51:07.839
Julie PEREIRA: So they had.

283
00:51:07.930 --> 00:51:23.970
Julie PEREIRA: They knew the how the program worked. But the kids did start the dividing the year we started evaluating. But for 35 other schools. It was their first years into the program. They just joined the the program when we came to evaluate.

284
00:51:24.710 --> 00:51:27.899
Julie PEREIRA: I don't know if that answers the first part of your question.

285
00:51:30.370 --> 00:51:38.580
She has to run to the microphone again. I'm sorry I was sitting on my chair again, so I'm back to my.

286
00:51:38.670 --> 00:51:49.800
you may think of, and running the analysis only with new schools. Perhaps that's more comparable to the controls. Just just a comment. But you answered the first question.

287
00:51:50.120 --> 00:51:54.830
Julie PEREIRA: okay, let me just note that down.

288
00:51:58.770 --> 00:52:06.130
Julie PEREIRA: okay. And then for the second question, that's in the the trick here to distinguish between

289
00:52:06.290 --> 00:52:30.989
Julie PEREIRA: the school effect and the program effect, because we we have no at the origin of the inside. The school of who is treated and who is not. What we are trying to do at the moment is gather national evaluation data for years before the program arrived in the school to see if we can assess a change in, in, in trend, in, in, in, in, in learning, in, in growth,

290
00:52:31.860 --> 00:52:36.190
Julie PEREIRA: But other than comparing those schools, and and

291
00:52:36.980 --> 00:52:50.969
Julie PEREIRA: and thinking that the public French system is rather unified, and because they are in the same neighborhoods, and because they are in in the same label. they should be similar enough to for comparing we we we

292
00:52:51.030 --> 00:52:58.580
Julie PEREIRA: we cannot rule out that this is a just different schools. I don't know if you, if you want to jump in on that? Or.

293
00:53:00.860 --> 00:53:02.490
okay, yeah, thank you.

294
00:53:02.550 --> 00:53:04.280
There are more questions.

295
00:53:06.560 --> 00:53:13.880
So yeah, I have a question.

296
00:53:15.080 --> 00:53:20.250
so how does it work in these schools. do they have an extra hour

297
00:53:20.390 --> 00:53:27.689
after school, or because otherwise negative effects can come about because they do something else. They don't do.

298
00:53:27.860 --> 00:53:41.629
Julie PEREIRA: Yeah, it. That's exactly that's exactly what we think. And so learning in in kindergarten, it's a little bit less formal. It's a bit less structured. So maybe that's also why we observe

299
00:53:41.890 --> 00:54:07.400
Julie PEREIRA: positive impacts in in kindergarten. But maybe that's also why those impact could disappear. when entering primary school. They don't have additional learning time, and they do have to get in and out of the classroom 3 times a week. The positive side that some teachers highlight is that the program enabled them to work with smaller groups of children, because, half of the class might be in violent

300
00:54:07.400 --> 00:54:18.710
Julie PEREIRA: classes. Wait, and why the others stay with the teachers. But that's the main, in my opinion, that's the main drawback of the program. It takes school time away from kids that are already disadvantaged.

301
00:54:20.350 --> 00:54:21.829
Okay, thank you.

302
00:54:23.130 --> 00:54:29.719
So I don't see any other questions. So thank you very much. What an interesting presentation.

303
00:54:30.210 --> 00:54:35.500
And we continue with the search presentations. It's also also online. How many of you

304
00:54:36.380 --> 00:54:37.230
with you

305
00:54:40.410 --> 00:54:43.829
and unmute yourself, please, so that we can hear you?

306
00:54:48.300 --> 00:54:54.320
Han Liu: Yeah, can you hear me now? Yes, we can hear you.

307
00:54:54.700 --> 00:54:59.730
We have. We have your we have to the screen with,

308
00:54:59.920 --> 00:55:23.480
Han Liu: it's also the other slides. The okay. Yeah. So thank you for all the help from. So yeah. my name is Han Leo from? I'm from the Department of Sociology at the University of this is the only of New York. The type of my presentation today is early, has conditions, family promises and education achievements.

309
00:55:23.570 --> 00:55:28.150
Han Liu: this is part of my dissertation research at your company.

310
00:55:28.970 --> 00:55:48.029
Han Liu: So in sociology, and also some other social sense discipline. early lifestyle has to be identified as the driving a factor of the production and also the reproduction of the free quality. most of the studies on this topic simple focus on early lifestyle and the consequence of your quality.

311
00:55:48.100 --> 00:55:55.740
Han Liu: but in recent decades. There's also increasing interest in the societal socioeconomic consequences of the.

312
00:55:55.770 --> 00:56:02.489
Han Liu: namely, how have a how our children's better attendant later in the life?

313
00:56:02.810 --> 00:56:19.959
Han Liu: And in the meantime, family is also constantly estimated. like family, is also constantly emphasized the in status of tenants. Research is the primary institution for reaching children, and is also a critical context for the development of children and their status attendants.

314
00:56:20.190 --> 00:56:46.290
Han Liu: And also currently, we're experiencing the second, the demographic translation. for example, there's a a constant decline in. And now now, by the fertility rate has been raising in research. That case So this changes in the family structure just to make it equally important for us to learn more about the role of family in this classification system.

315
00:56:47.400 --> 00:57:16.429
Han Liu: and also according to some studies early, that crafts may also affect public processes and home the environment This funding from prior research suggests that from the dynamics may at least the part of the impact of fully left out that as a terminus outcome how I work this important mechanism has been directly tested in Imperial research. in this study, I

316
00:57:17.050 --> 00:57:34.280
Han Liu: I try to test the home environments mediate of the price the impacts of on educational chambers and assignment. I focus on education because it's a key factor for social mobility. in stratification, research and statistical models.

317
00:57:34.370 --> 00:57:43.320
Han Liu: And also I'm using prospective data reported by mothers to capture the live class conditions and also home environments. Early in life.

318
00:57:44.640 --> 00:57:53.229
Han Liu: about the link between childhood and education plus studies, have I processed the 2 potential mechanisms.

319
00:57:53.590 --> 00:58:02.960
Han Liu: The first one is a fixed, a lot of Co mechanism. all right, I may affect the children's free and the body development. And

320
00:58:02.970 --> 00:58:27.340
Han Liu: The impact on this development processes may have. we have some for the impact on their education. And the second, the mechanism is fecal social Early life, perhaps, may also affect the children's educational situation. and according to the Wisconsin model of that is determined situation is actually very important, the first of final status attendant.

321
00:58:28.020 --> 00:58:43.509
Han Liu: there's 2 passwords both critics. Next, healing paths for the lifestyle on education. And this next few impacts have been supported by research focusing on both ways current conditions and the retrospective childhood.

322
00:58:43.530 --> 00:58:55.510
Han Liu: however, there are also some limitations in prior research at this topic. first, that again, I to create attention to the data action with the children and social institutions to interact with

323
00:58:55.520 --> 00:59:02.930
Han Liu: for children before school age, family is the most important social in social institutions, that kind of factor.

324
00:59:02.940 --> 00:59:27.569
Han Liu: And this, first, the limitation is partly because of second part I'm talking about here. is a data limitations for studies on this topic. When the family based on retrospective data, therefore they may not be able to use them. we find the measures for home environments, or as the family level to test this important mechanism.

325
00:59:28.800 --> 00:59:32.890
Han Liu: so why should we care about family?

326
00:59:33.650 --> 01:00:01.440
Han Liu: hey, I want to emphasize 2 potential mechanisms first, children's health issues may have some negative may have some negative impact on the family as a whole Children with health issues may have some special may have some special kind of need, and it's a special has. Currently it can lead to financial costs that can be used to otherwise. to improve children's education, and also to have Chinese may affect the parents, labour market behavior

327
01:00:01.780 --> 01:00:20.990
Han Liu: and also children's health conditions can affect the parent's sex psychological house, and also sometimes to have the useful stuff to come and test, and the calendar of social isolation of the family posted psychosocial processes can have some negative impacts on the children's while.

328
01:00:21.330 --> 01:00:43.739
Han Liu: And in addition to this, impact on the firm as a whole children's health issues may also affect how Paris, Tristan, from an economic perspective. If players want to maximize the intergenerational transmission of the ses It may not be the best choice for them to invest in the education of children which have issues.

329
01:00:44.100 --> 01:00:59.579
Han Liu: and also on the other hand, it's possible that parents may want to compensate for all the live adversities. So in this case took over which houses may actually receive more, your investment and support from their parents.

330
01:01:01.080 --> 01:01:27.830
Han Liu: so based on the above discussion, I actually have a pair of competing hypotheses first, to just have problems make consumers to labor market participation may lead to from the dysfunction a. And he would just like to be manage the investments and support from heirs. therefore, home environments may actually extensively in active impacts of all that kind of

331
01:01:27.970 --> 01:01:30.110
Han Liu: children's educational outcome.

332
01:01:30.150 --> 01:01:43.669
Han Liu: however, because children have problems, they also need to compensatory parenting strategies. it's also possible that chance would have problem actually associated with some

333
01:01:43.780 --> 01:01:50.639
Han Liu: with some improved for me, right on that that can actually the price, the detrimental impact of

334
01:01:50.840 --> 01:01:52.890
Han Liu: you shoes.

335
01:01:54.050 --> 01:02:14.999
Han Liu: So this, I'm a hypothetical. And to test this pair of competing hypotheses, I'm using data from the injection, the national logic to no sort of use. 1,979 from the United States. So this is a nationally representative sample of people who were 14 to 21 years of age on December 30, first 1,978,

336
01:02:15.040 --> 01:02:18.919
Han Liu: and since the seventeen-nine survey had to be

337
01:02:18.960 --> 01:02:27.530
Han Liu: a PIN of project focusing on children, the young dolls up on to the will be the original.

338
01:02:27.670 --> 01:02:39.790
Han Liu: This study you, this 17 wave of data from 1,986, the first wave of the child and the others project to 2,018. In total. They are.

339
01:02:39.960 --> 01:02:46.030
Han Liu: There are 11,545 children covered in this.

340
01:02:46.170 --> 01:03:02.369
Han Liu: So the the I'm using here have 2 streams that I want to am set for stays. Prospective data with the information I'm charged with has for the environments, and it also has which information on education. this is the key variable I am using with this study.

341
01:03:02.840 --> 01:03:19.029
Han Liu: and also in this, that the children can be linked to their mother's. so it's possible for me to control for some imported the confounders. like masters prenatal behaviors. this, this, matters are usually not available. We need retrospective data

342
01:03:20.520 --> 01:03:37.560
Han Liu: and moving on to variables. The outcome I'm looking at is education. I measured it with. 2 sets of variables. The first side of variables are from the people in the way to achievement test. this is a test of math.

343
01:03:37.640 --> 01:03:56.209
Han Liu: yeah, it's a test of my time greeting abilities. I it has high. It's it's it's really comprehension.

344
01:03:56.390 --> 01:04:08.710
Han Liu: And the last 2 people that variable I'm using is about their education in the formal system. If it's whether or not the child attendees college by age 20,

345
01:04:10.240 --> 01:04:19.559
Han Liu: and about

346
01:04:19.590 --> 01:04:25.550
Han Liu: the first one. It's a low but weight defined as

347
01:04:25.790 --> 01:04:27.560
Han Liu: and the second

348
01:04:27.620 --> 01:04:42.499
Han Liu: on the laptop measure. It's a number of unices requiring medical attention or treatment in the 12 months preceding each wave of the survey. In this study I'm using the average number of unions we both reported between 8 0 and 8 5.

349
01:04:42.940 --> 01:04:58.309
Han Liu: And the the the mediator I'm using is this study. It's about home environments and myriad with for motivation management of the environmental inventory. there's a total score for the children's home environment.

350
01:04:58.460 --> 01:05:04.650
Han Liu: and there are also separate scores for cognitive accumulation and the emotional support at home.

351
01:05:05.000 --> 01:05:08.019
Han Liu: And also in this study I control for the

352
01:05:08.200 --> 01:05:28.249
Han Liu: so the demographic of of the child and the manager. And also I control for it is. And also

353
01:05:29.560 --> 01:05:49.509
Han Liu: about my analytics. Touch date. I test my hypothesis within a mediation analysis framework. Specifically, I'm using the Php message to test the mediating effect of from the environment this method adjusts for discounting issues in nonlinear capability models.

354
01:05:49.520 --> 01:06:00.429
Han Liu: one of my business variables about the college attendance at 820 is binary, and I view the logistical regression to estimated as it.

355
01:06:00.560 --> 01:06:08.049
Han Liu: there's no linear models. I have calendar issues in mediation. And I's the case. We message address for this

356
01:06:08.270 --> 01:06:18.989
Han Liu: and also this method, a lot for multiple mediators. So I can test the the separate effects of cognitive cognitive simulation and the emotion of the product home.

357
01:06:19.290 --> 01:06:32.449
Han Liu: about my analytics sample. First, I I constrained my analysis to children, for after 1,986 the first wave of the survey so that I have data for all their left calls.

358
01:06:32.800 --> 01:06:37.409
Han Liu: and also for the

359
01:06:37.440 --> 01:06:50.600
Han Liu: we to extend 814 I just to to to work 14 years or older in 2,014 that's because the test the the math and reading had. what's not

360
01:06:50.830 --> 01:06:53.560
Han Liu: what? No longer available after 2,014,

361
01:06:53.610 --> 01:07:07.740
Han Liu: and for the analysis on call. If a sentence at age 20, I I restrict my to respond to what 20 years are older than in 2,018 the sample set is actually similar for this 2 set of analysis.

362
01:07:08.460 --> 01:07:38.340
Han Liu: And because I'm using data up to 20 years about half of to bring in my sample, actually have meeting values To preserve this case, I it's much for implication. specifically, I use the to level with the fax multi detail for the imputation. it's a lot for varying intercepts across individuals and families that can capture a variation at the individual level, and also when we reach out to the family level.

363
01:07:39.520 --> 01:07:46.670
Han Liu: So here I may recall. let's start from the effects of what we, our test scores.

364
01:07:46.720 --> 01:08:04.750
Han Liu: So for you to come, I estimated 2 models that reduce the model. It's the model with us, the measure of a home environment. This gave us the total effect of the low of the health issues pay a local rate And then I estimated a full model.

365
01:08:04.750 --> 01:08:20.949
Han Liu: It has both the Canadian variable, local retail, and also from the environment. And the coefficient of low right here is the the effect of with after home environments are controlled. so according to the results here.

366
01:08:20.950 --> 01:08:30.030
Han Liu: for all 3 has the scores. The total effect of low birth rate is negative, and sleep exercise is actually similar across outcomes.

367
01:08:30.140 --> 01:08:40.540
Han Liu: for only if we add home environment into the model the effects of is no longer significant and the effect of a home environment with

368
01:08:40.550 --> 01:08:48.389
Han Liu: significant and the positive, which means home environment. can you prove? Or it's possible to which has the scores

369
01:08:48.490 --> 01:08:57.120
Han Liu: So this results for just that home environment can explain the negative impact of low birth rate of past this course.

370
01:08:57.850 --> 01:09:01.399
Han Liu: And in the next step I put the

371
01:09:01.450 --> 01:09:18.379
Han Liu: he composed the mediating impact of home environment into his 2 components, cognitive simulation and the emotional support, and according to the results the effects from both. His 2 components are much more significant, and they are both active.

372
01:09:19.610 --> 01:09:23.739
Han Liu: So it's not a for

373
01:09:24.600 --> 01:09:40.439
Han Liu: and moving on to the effects of your mes shoulders test the scoff between each 10 at day to 14 So, according to the red deals, the model of the total effect of the you're on, it's it's before it's actually close to you.

374
01:09:40.440 --> 01:10:00.089
Han Liu: And again, this positive effects disappear after we add from the environment into the model. and the significant. The the column here shows that the difference between the reduced model and the full model is significant. just you that the mediating effects of from the environment is significant

375
01:10:00.090 --> 01:10:14.400
Han Liu: and also according to the bottom panel here. if we decompose the mediation, you better for the environment into the 2 components they both contribute to this mediating process.

376
01:10:16.460 --> 01:10:43.929
Han Liu: And last we look at the effects of both ways and unices. college attendance at x 20 according to the men meditation analysis the effect of for me what the immediate need for us for me, environment is now significant for college attendance no matter how we better or the like house. However, if we decompose home environment into these 2 components.

377
01:10:44.190 --> 01:10:54.389
Han Liu: we can't. if we look at the depend on the right the vacation or cognitive simulation is actually positive and much know is significant.

378
01:10:54.420 --> 01:11:01.160
Han Liu: and the mediation through the motion of the party is actually active and significant, which means.

379
01:11:01.660 --> 01:11:08.609
Han Liu: Yogi. It has some an active impact on college attendance through home environments.

380
01:11:10.530 --> 01:11:38.850
Han Liu: so we have some conclusions and some explanations for the funding first, about the fax of fully like kind of our education. So, in fact, actually depends on the timing of the nature of the test problems we are looking at for locals way, it has negative impacts on the types of scores. And it's negative effects can be totally explained by home environments, because when you dig up, the effect is no longer significant.

381
01:11:38.890 --> 01:11:40.060
Han Liu: dude.

382
01:11:40.080 --> 01:11:48.409
Han Liu: This findings suggest that low possibilities leading to disadvantage of environments with negative impacts on children's education.

383
01:11:48.540 --> 01:12:14.350
Han Liu: And also if we look at the on this before it's it's it's actually you have positive effects on test scores this funding is actually different from what I found from local to it. And it's also different from funding from most of her research. I think this difference is because of how I minor.

384
01:12:14.370 --> 01:12:19.350
Han Liu: It includes both the bay and mild health conditions and become.

385
01:12:19.440 --> 01:12:31.910
Han Liu: And because, may, I have issues, I just to more prevalent my children most of the units is captured that this variable may not be safe enough to have them better.

386
01:12:32.080 --> 01:12:43.140
Han Liu: That's not going to impact on children's education or development. however this may or the issues may be, or fact, how with the children and we still have some.

387
01:12:43.360 --> 01:12:51.709
Han Liu: we still have some impact on their phone environment. As therefore, the the only things may actually lead to some positive effects.

388
01:12:51.750 --> 01:12:53.690
Han Liu: children

389
01:12:54.040 --> 01:13:07.150
Han Liu: performance in in Princess Test. And this person, in fact, actually planned by home environment services supporting The explanation through family processes.

390
01:13:08.400 --> 01:13:28.399
Han Liu: And another major funding is that home environments also have different damages specifically in this study attached to cognitive team relation and the motion of support cognitive team relation is actually positively associated with both test scores and call attendance. how I way, in terms of the emotional support.

391
01:13:28.400 --> 01:13:45.830
Han Liu: also, it's still positive associated with test scores. it's actually in that field associated with called attendance at 8 20 So this funding suggests that colleagues but call call it attendance and call it the Commission actually requests more to test the scores.

392
01:13:46.050 --> 01:13:47.470
Han Liu: And

393
01:13:47.720 --> 01:14:09.879
Han Liu: and interestingly, I also found that the emotional support can have some networks on courage of the mission especially for children who get the emotional support because of the aircraft conditions. it's possible that when this happens, there may be some negative impacts that are not captured at the home environment measures I'm using?

394
01:14:09.930 --> 01:14:18.420
Han Liu: so future research, should I look more into this issue back in terms of leading to to that same impact

395
01:14:18.520 --> 01:14:38.519
Han Liu: and also to have some potential to funding issues with them. Who some like logical factors that make children require more emotional support. may. Also they feel they're likely without call if the mission and my study may not be able to address all of this confounding practice.

396
01:14:39.720 --> 01:14:48.399
Han Liu: and the United States also has some limitations first, my union's.

397
01:14:48.450 --> 01:15:13.080
Han Liu: and also I do not have later such a the out education, because most of most of the the outdoor in the. So we are still in there early adults. and also the the only thing is measured captured in the survey.

398
01:15:13.370 --> 01:15:19.689
Han Liu: just all of my funding. thank you for for your attention.

399
01:15:21.260 --> 01:15:22.219
Thank you.

400
01:15:27.170 --> 01:15:28.130
Okay.

401
01:15:28.350 --> 01:15:30.989
so we have time for questions.

402
01:15:35.070 --> 01:15:37.990
Okay, I I can start

403
01:15:38.780 --> 01:15:42.609
I actually, I based on your introduction.

404
01:15:42.710 --> 01:15:58.539
I expected that you would look at an interaction effect of the environment and early health, so that you suggested that the effect of early health on achievement would depend on

405
01:15:58.580 --> 01:16:03.259
the family environment. You could also interpret that as Ok, there is basically.

406
01:16:03.590 --> 01:16:18.390
for example, a negative effect, but it is stronger in certain families, and so weaker than other families. Did you think of modeling it that way, or or is that? Or do you think that it is actually more a mediation problem

407
01:16:19.790 --> 01:16:30.640
Han Liu: oh, central for this. Oh, I think it's a very good part yeah, I think, the the fax of childhood.

408
01:16:30.720 --> 01:16:59.580
Han Liu: Of course it can be for a short social classes or other like straight offs of inequality. So I think, if I do a moderation analysis. There may also be some interesting funding but the specific focus of this study is about whether or not this but they have the issues kind of back for me, environment which for the affects the education. So it's just about my focus here. I think it can happen both ways.

409
01:16:59.970 --> 01:17:01.879
Han Liu: I the focus great on.

410
01:17:02.680 --> 01:17:03.550
Okay.

411
01:17:04.900 --> 01:17:10.180
Karl, I see that you've unmuted yourself. I think you did so because you want to ask a question.

412
01:17:10.290 --> 01:17:19.540
Carlo BARONE: yeah. concerning the relationship between low birth weight and home environment.

413
01:17:19.680 --> 01:17:37.809
Carlo BARONE:  do you see the home environment as a mediator or as a co-founder? Meaning. Do you assume that? A low birth weight reduces the home support for the children, the mediator interpretation.

414
01:17:38.170 --> 01:17:50.860
Carlo BARONE: or that it actually the home environments? is a proxy for a measured home coast, the low birth weight.

415
01:17:51.370 --> 01:18:11.129
Han Liu:  yeah. So it's it's also a good phone So my survey is low can lead to disadvantage from requirements. And in in my models, I actually controlled for this of the economic conditions of the family extra time to a phone

416
01:18:11.130 --> 01:18:35.120
Han Liu: including the mother's education occupational prestige and the income. So I can choose for the socioeconomic factors that can potentially affect both house and the home environment, but of call, that kind of control everything. So there can be some potential confounders that can tribute to both vectors. So it's definitely possible.

417
01:18:36.770 --> 01:18:43.799
Okay, thank you. Could I follow up on this? Because I think some of these women will have 2 children.

418
01:18:44.100 --> 01:18:57.110
and our second child may actually have the the health problems. And then a way to maybe solve this issue is to see whether they give more emotional support to their first child, so it does not have the low birth weight.

419
01:18:57.270 --> 01:19:04.639
They give less support to the second child or the other way around. If the first child has the Albert birth weight. And

420
01:19:04.810 --> 01:19:07.110
so within family differences.

421
01:19:07.180 --> 01:19:28.639
Han Liu: Yeah, yeah, so yeah, I actually thought about this potential like, fixed defects, models to to address this issue but as I get to deals to in my theory, part actually, some of the the IP of maxing zones I'm trying to calculate. It's kind of the issues may affect the whole family, which kind of in turn affects all

422
01:19:28.750 --> 01:19:56.129
Han Liu: children in the family, no matter how many children they are, and for within family differences I I kind of over generous to funding to this potential like mechanisms to with the whole family. So that's where I'm not using it. But I think it's a good part to kind of try to do some such. It can be tested with quick defense models so that I can know what's happened is this co-family processes can choose for.

423
01:19:56.440 --> 01:19:58.030
Han Liu: Yeah, it's about a good pond.

424
01:20:01.250 --> 01:20:04.820
More questions, either in the room

425
01:20:04.980 --> 01:20:06.240
for online.

426
01:20:10.900 --> 01:20:20.750
I I do have another one. Maybe you touched upon it, and I didn't hear it. That's possible, because the sound is not so good here. But

427
01:20:21.030 --> 01:20:23.530
would you agree?

428
01:20:23.770 --> 01:20:25.010
that's

429
01:20:25.300 --> 01:20:44.619
the difference between the the low birth rate and the illnesses? could have something to do with the fact that the low birth rate is not so visible. so they, these parents, may actually not see that as a health problem. But they might just think, oh, my, this child is not so smart.

430
01:20:44.720 --> 01:20:50.379
whereas the illnesses are very visible, and may therefore

431
01:20:50.690 --> 01:20:54.759
lead to way more support from the parents.

432
01:20:55.660 --> 01:21:20.159
Han Liu: Well, yeah, yeah, I think it's totally possible. Yeah, especially in nowadays. low may not need to extremely have issues in. They feel. in the first a few years. So it's totally possible that parents do not treat it as a perhaps problem. But we just fix it. That has a problem. So it's it's totally possible, I think.

433
01:21:21.560 --> 01:21:23.190
Okay, thank you.

434
01:21:23.760 --> 01:21:28.089
So if there's no more questions, then I think we can go to the

435
01:21:28.170 --> 01:21:32.579
last presentation, and that's agree on the onsides presentation.

436
01:21:33.470 --> 01:21:40.690
So we have to do all kind of things now. But I think you'll need to share screen. Yes.

437
01:21:41.100 --> 01:21:42.440
and then

438
01:21:43.110 --> 01:21:44.100
so

439
01:21:45.190 --> 01:21:45.940
yep.

440
01:21:49.560 --> 01:21:53.439
okay, let's hope that it works perfect. Okay.

441
01:21:53.790 --> 01:21:55.260
let me

442
01:21:56.470 --> 01:21:57.400
perfect.

443
01:21:57.760 --> 01:21:59.500
Okay,

444
01:21:59.940 --> 01:22:01.850
this is how I know.

445
01:22:03.890 --> 01:22:18.449
yes, I was checking. Okay, good morning. Everyone. My name is Valentina, and I will present some preliminary results from a part of a broader project that I'm working on, together with my colleague, Kathryn Bark, who couldn't be here today.

446
01:22:18.470 --> 01:22:31.930
So this is an ecrc funded project that looks at social and ethnic biases in primary education. And today, what I'm presenting has to do with the Scs gradient in teacher bias. So teacher perception of academic abilities

447
01:22:31.990 --> 01:22:39.219
of primary school students in England, Scotland, and Germany, and a preliminary investigation of some of the potential mediators.

448
01:22:40.170 --> 01:22:50.089
Okay, so a little bit about the theoretical background. So the overall motivation for the project of a role. And this study is

449
01:22:50.380 --> 01:22:54.019
fits and the idea of the self fulfilling prophecies and feedback loops.

450
01:22:54.030 --> 01:23:07.710
So the idea that teacher, feedback, input output and overall even the quality and the type of teacher. Student relationships are affected by teacher expectations and perceptions about students, not only abilities but also attitudes.

451
01:23:07.890 --> 01:23:22.119
and then in turns, teacher, feedback, input output and teacher. Student relationship affects student attainment, student effort and attitudes and student self-image, which then feedback into teacher expectations and perception.

452
01:23:22.570 --> 01:23:35.920
So exactly where we are in terms of our focus is this idea of teacher expectations and perceptions, and whether they're biased and we'll get to, what do we mean by bias? And how should this word be taken in the context of this project?

453
01:23:36.760 --> 01:23:56.769
So when we look at teacher judgment and teacher bias, the idea is that the accuracy of teacher. Judgment varies. So we know from a variety of different empirical studies that the shared variance variance between teacher judgment and student achievement is around 40%. So that leaves another 60%

454
01:23:56.770 --> 01:24:08.119
And this remaining variance is what we could think of as the inaccuracy. And it could be positively or negatively biased, so it could be an over and underestimation

455
01:24:08.120 --> 01:24:31.090
of students achievement. There is a variety of empirical evidence that shows that students from more socioeconomically disadvantaged families, but also from minority background and in different cases from different genders, often face lower teacher. Expectations is a either objective achievement, so this means that their abilities are over or underestimated by teachers

456
01:24:32.050 --> 01:24:38.609
in terms of the overall framework. If we think of teacher judgment somewhere on a spectrum between.

457
01:24:38.650 --> 01:24:48.870
let's call it perfect information based judgment on one hand, and on the other side, automatic judgment guided by stereotypes. There are a variety of ways that we can

458
01:24:49.170 --> 01:25:09.860
think how socioeconomic background might impact where the judgment stands on this spectrum. So we can, of course, things of think of burgl culture, reproduction theory and look at teachers as gatekeepers and school. It's places where specific kinds of culture are rewarded and and and

459
01:25:09.930 --> 01:25:23.410
judge more favorably by teachers. So the idea is that teacher. Assessment of student performance is not only based on students ability, but also from students and their families. Cultural capital. And this could be expressed

460
01:25:23.410 --> 01:25:40.959
and manifest in a variety of ways, as we know, for example, linguistic aptitude, anything from mannerism and accents to extracurricular activities of the child and the child and the family together, and parental support. So parents who contact parents interest in child's education, and so on, and so forth.

461
01:25:41.570 --> 01:26:07.910
a slightly different way to look at it, but not necessarily opposite, is to look at correspondence theory, and to think of school systems and school themselves as to correspond to labour market and workplaces. So teachers, like employers would reward certain non-cognitive skills depending on the roles and position of a student, or or exactly as employers might do with employees. So the idea here is that once again, when teacher form judgments about

462
01:26:07.910 --> 01:26:34.090
children's ability, they are actually also taking into account certain non-cognitive skills and the type of non-cognitive skills that are taken into account might be different across the Ses background. So, for example, low ses students might be rewarded for certain trades like obedience and compliance with homework completion, whilst high ses students might be rewarded for independent thinking and creativity.

463
01:26:35.950 --> 01:26:51.889
But this is not to say that there are not other factors. The don effect where teacher judgment stands on that spectrum. And these are factors that overall guide us in the cross country. Comparison that we're trying to carry in this project and in this

464
01:26:52.140 --> 01:27:12.850
investigation in particular. So there is. Of course, there are different levels, one of them, the first one is teacher personal characteristics. And here we see attitudes, knowledge and mindset. But of course, we're also talking about personal characteristics. So their own ses background of the teacher, gender, age, ethnicity, and so on. And then we have institutional context.

465
01:27:12.850 --> 01:27:27.240
So this has to do with the conditional regulations on school. So from tracking to the degree of standardization, the type of testing that is done in school, and also things like teacher training, which you know can be very important.

466
01:27:27.280 --> 01:27:38.330
And then we have situational characteristics in the specific school and in the classroom. Even so, time, pressure, judgment goals. And then the type of social cues that teacher can use 101

467
01:27:38.940 --> 01:27:58.689
to form opinions. So, broadly speaking, the framework that links social class to teacher, assessment of language goes through skills and and ability in general. And the idea here is that there are certain sets of mediators that mediate the relationship between social class and teacher assessment.

468
01:27:58.690 --> 01:28:11.640
And broadly, we're talking about 3 sets of mediators. If we want a child's behavior and social emotional issues, child's attitudes towards school learning and the teacher themselves, and parental involvement in child's education.

469
01:28:12.900 --> 01:28:25.970
so we have quite a bit of information about data which I won't go into details. But it's there if you have questions, because working across 3 different countries with 3 different, the sets of data triplicates, the amount of information.

470
01:28:25.980 --> 01:28:44.270
So we're talking about England, Scotland, and Germany. So we're using the millennium cohorts study in England growing up in Scotland for Scotland and the national Educational panel study for Germany. These are birth cohort that don't exactly match, but they roughly refer to children born at the beginning of the century.

471
01:28:44.860 --> 01:28:45.970
The millennium

472
01:28:47.000 --> 01:28:57.480
so again won't go into details. But our main, like target time, is the end of primary school so roughly when children are at 10

473
01:28:57.530 --> 01:29:19.909
and then the we have 2 other time points at the start and during primary school. So the match across countries in terms of age is not perfect, but our primary focus is that children are evaluated as close as possible to the end of primary school the last year with the in the primary school system, and then the 2 time points before that. There's

474
01:29:20.000 --> 01:29:23.399
We're ready in primary school. So not in kindergarten or preschool.

475
01:29:24.570 --> 01:29:53.849
so I won't go through all of the variables, but just to say that when we then I will then show you how we capture teacher by yes, or in accuracy in teacher judgment. But the main 2 variables that we are looking at is the language, skill, test, and the teacher rating of child's language and skills. So for the language skills tests, these are different cognitive, language-based tests administering the 3 countries with the survey. And then we have teacher reported rating of the child skills.

476
01:29:53.910 --> 01:29:56.200
and these are both at the same time point.

477
01:29:56.280 --> 01:30:06.870
We then have a measure of ses which in our case it's income quintiles, a variety of prior ability that our language-based skill test or hydrocognitive skills test.

478
01:30:07.280 --> 01:30:24.300
And then for our sets of mediators, child behaviors and other social emotional issues are captured through the user. Validated scales mostly strengthen difficulty questionnaires up scales, and some of the big fives. We have child attitudes who are at school learning and teachers which are broadly

479
01:30:24.850 --> 01:30:45.679
 articulated around the concepts of effort, school enjoyment, academic self concept and teacher student relationship. And then parental involvement was split into 2 parts, a school base and a home based under the assumption that school-based parental involvement is what's most visible to the teachers themselves.

480
01:30:45.780 --> 01:31:11.660
So for school-based parental involvement. We have participation in Parents evening, and then especially other specially arranged meetings, and whether they were arranged by the parents themselves or not. So here the idea of whether they are reactive or proactive as a form of parental involvement, and then the volunteering of parents in a variety of extra activities like so curriculum activities from school board, fundraising day away and so on.

481
01:31:12.270 --> 01:31:29.589
And then for home-based parental involvement. we're looking at parents interest in childhood life. So whether the parents ask questions to the child, and how often about what's going on in school, parental, educational aspirations and opinions on education in general, and the value of education qualifications. And so on.

482
01:31:30.420 --> 01:31:42.300
Okay, so our analytical strategy is at the first step to what we call estimate residual. So this is how we're capturing, whether teacher perceptions are accurate or inaccurate, so whether they are negatively or positively biased.

483
01:31:42.320 --> 01:31:47.850
then we have univariate analysis on whether there is an Ss gradient. And finally, we get to the

484
01:31:48.150 --> 01:32:14.140
to the mediation part. So at this stage, because this is preliminary results. What we're presenting is a very simple, value-added wells framework. So we're just comparing the role model where ses so income quintiles is the only predictor of teacher bias to models where one by one we insert the mediators, and we check if the difference in the sees gradient is significant or not, and by how much is reduced.

485
01:32:14.820 --> 01:32:36.250
Okay, so this is the idea of why we use residuals to capture teacher bias. And it's a lot simpler than it might look. So the idea is that, given a certain student test performance X, one. For example, you would expect, based on the the data that we have and the model that we run a certain teacher assessment

486
01:32:36.360 --> 01:32:45.570
y. One. But in reality what we get is z. One. So this residual of the regression captures whether the

487
01:32:45.570 --> 01:33:14.459
teacher is accurate or not, and by how much and whether the accuracy, the inaccuracy sorry, is positive or negative, so a positive residual would indicate that the teacher is overestimating students, not attitude, but abilities, and a negative residuals would signify the opposite. So residuals are then standardized, but because by definition they have a mean of 0. Demeaning them doesn't change the sign. So it doesn't change the substantial interpretation of what a negative or a positive residual means.

488
01:33:15.080 --> 01:33:18.150
Okay, so we had a bit of descriptors. But I'll skip them.

489
01:33:18.290 --> 01:33:31.830
So these are the models that we use to estimate teacher bias. You might notice that there are a few more than just the one predictor. So the main predictor is the test score in language skills at time 2 for all 3 countries.

490
01:33:31.830 --> 01:33:55.919
We then also control for age of time, of testing, and for a few prior language skills measure one or 2 depending on what is available in the countries and other cognitive abilities. These other cognitive abilities measures have to do mostly with non verbal reasoning or spatial reasoning. So here the idea of what we're doing this is that there is, of course, an underlying question of whether it is the test that is inaccurate

491
01:33:55.970 --> 01:34:15.359
or is a teacher that is an accurate. So is it actually what what we see as bias is still always possible that it's just because teachers see more, understand more of the actual abilities of the child that one singular test could capture. So the idea of controlling from some not just language, but non-language

492
01:34:15.400 --> 01:34:28.389
skills, is to kind of weed out as much as possible. This. So what we see is that in all 3 countries there is still even in England, which is the the one with the highest share of teacher

493
01:34:28.540 --> 01:34:39.880
judgment that is explained by our model around 30%. In all 3 countries there is a fair amount of variation in teacher assessment that doesn't seem to be explained by

494
01:34:40.140 --> 01:34:44.460
test scores, by cognitive or other cognitive abilities.

495
01:34:44.760 --> 01:34:52.959
There are differences across the 3 countries. In particular, Scotland has a much lower r squared or a 13%.

496
01:34:53.000 --> 01:35:08.589
Scotland has a slightly different wording of the question that it's used to us teacher, to rate children. So we're still kind of debating how to interpret this difference. If it's substantial difference, or it's a product of the difference in measurement.

497
01:35:09.530 --> 01:35:14.619
Ok, when we look at the Asex gradient just by looking at incon Quintiles on its own.

498
01:35:14.660 --> 01:35:42.080
Very interesting to see is that doesn't seem to be a different across the country no recognizable pattern. But in all 3 countries the Ascs gradient is extremely visible. So we're talking about the lowest children in the 2 lowest Quintiles having their academic abilities in language skills underestimated by teacher in a statistically significant way, and the opposite being true for children at the top of the distributions of the 2 highest quintiles.

499
01:35:42.770 --> 01:35:43.630
Sorry.

500
01:35:46.320 --> 01:36:12.540
Okay, so I won't present all of the models. But I have them in the appendix. I'm welcome to discuss in more details with you, basically, what I'm showing here is results for this very preliminary preliminary investigation of the mediators of this ses gradient. So the black bar. This is what we call the role. Model. So it's just income Quintiles predicting

501
01:36:12.790 --> 01:36:14.110
teacher bias.

502
01:36:14.270 --> 01:36:19.940
And we take the Ss gap as the difference between the bottom and the highest queen tile.

503
01:36:20.000 --> 01:36:46.870
Then, in the subsequent models. What we're doing is that we're doing one mediators set of mediators at a time, and we can see that in almost all cases the reduction in the ses gap is significant, not in all cases, but in almost all cases, but the signs of the reaction, the the reduction. Sorry differs across mediators, but also across countries. So this is where

504
01:36:47.170 --> 01:36:55.429
there is, and there are interesting results which are still very much debating. So I'm very interested in your opinion. So what we can see is that looking at England.

505
01:36:55.470 --> 01:37:24.579
the the brown bar which is the behavioral issue and social emotional skills, the reduction in the Scs gap. Here is the largest. So around 50% of this reduction of the Ses gap seems to be mediated by behavioral issues and social emotional skills we then have around 20%. And again, these are a separate model, one mediation at the time. It's we're still not at the contribution of different sets of mediators?

506
01:37:24.740 --> 01:37:32.540
so 20% when we look at child attitudes towards schools. So again, this is after school enjoyment and

507
01:37:32.550 --> 01:37:45.930
self-.

508
01:37:46.040 --> 01:37:54.319
When we look at Scotland. It still seems that the first 2 let's say, sets of mediators are the most important.

509
01:37:54.330 --> 01:38:12.089
but this time it it doesn't look like when we look at the first model. So the one with behavioral issues and social emotional skill that the mediation is a lot larger than child's attitudes. So both models have around a reduction of around 30% of the initial ses gap.

510
01:38:12.280 --> 01:38:25.030
And then for the parental involvement on its home-based parental involvement. significantly mediates the the ses gap.

511
01:38:25.070 --> 01:38:29.459
of of a lot more than in the case of England.

512
01:38:29.620 --> 01:38:53.670
The results for Germany are quite different, which is very interesting, and not too far off from some of our expectations. So, even though all of the models that we could run, we have no information about teacher student relationship. But in all of the model that we could run there is a reduction in the ses Gap. That is statistically significant. The sizes of these reductions are different. So in Germany it seems very much

513
01:38:53.670 --> 01:39:07.099
that the home based parental involvement is by far the largest mediators, and if we were to look into the actual model, we would see that one of the main drivers of this is parental educational expectations.

514
01:39:07.480 --> 01:39:08.370
which

515
01:39:08.730 --> 01:39:27.370
kind of tracks with our ideas of what's going on in Germany. So get some conclusions. In all 3 countries, teacher assessments. appears to be only partially accurate. At most, 30% of its variation is explained by language skills and prior language skills and and other cognitive abilities.

516
01:39:27.540 --> 01:39:34.309
 teacher, assessment is also strongly graded across along the Ss lines.

517
01:39:34.400 --> 01:39:51.970
and the model mediators account for even altogether. At most 45 to 63% of the Ses gap. But there are differences across countries, and there is still a part of the Scs Gap that doesn't appear to be mediated by any of the mediators that we modeled.

518
01:39:53.350 --> 01:40:11.650
I put on a few discussion points. I just wanted to mention that. Of course, this is still in the preliminary stages. We're still discussing because when you compare different countries and you're trying to harmonize data that was not created for the purpose of being harmonized. There are a lot of discussions about, are we actually using

519
01:40:11.760 --> 01:40:27.339
measures that are comparable across countries. So I think, in the, in the first step, when we estimated residuals the different results for Scotland, they kind of bring into questions whether the how we can interpret that difference, because the question asked in the

520
01:40:27.530 --> 01:40:50.089
Scottish Survey reflected the overall curriculum for excellent that is used in Scotland. That is, a well-rounded and broader approach, where teachers were asked first to report at what level the child was working, and then within that level, at what kind of level of proficiency. So there is a slightly different approach to it.

521
01:40:50.180 --> 01:40:53.910
And then, in general, in terms of why? We're seeing

522
01:40:54.000 --> 01:41:09.690
differences in terms of mediation, and which mediation tends to be possible, it tends to be more important. we've been thinking about, of course, not. Just differences in school system, for example, does the teacher change every year. What is the general one?

523
01:41:09.800 --> 01:41:29.520
focus that teacher are asked to keep in mind when evaluating children's abilities, but also what are salient science, or or it tells of the ses, background, or things that teachers might take as more important when assessing children's abilities.

524
01:41:29.620 --> 01:41:53.929
So yeah, next steps, we have a bit more fine-tuning to do with weight that needs to be constructed. To account mostly for non-matching or not consent to the teacher's survey, and then we want to explore some factor, component analysis, to explore the different mediations, but also as a method of dimension reduction. To then do some different mediation analysis, more sophisticated

525
01:41:54.060 --> 01:42:04.359
stuff. So I'll just leave you. Thank you very much. And this is the project website. And if you're interested, you can just type by B project, and it will come up and there is more material. Thank you.

526
01:42:09.830 --> 01:42:10.620
Sorry.

527
01:42:10.760 --> 01:42:17.610
Thank you.

528
01:42:18.530 --> 01:42:21.530
There we are. Yeah.

529
01:42:26.230 --> 01:42:41.089
Hi, thank you very much. It's very interesting. I have a one question. and it might be just that I just missed it because, there was a lot of information. It was. yeah. the teacher assessment?

530
01:42:41.500 --> 01:42:47.070
Is it part of the it's a teacher assessment. That's part of the school

531
01:42:47.150 --> 01:42:50.739
system. It's or is it for a survey?

532
01:42:50.750 --> 01:43:13.449
It's for the survey. So in the context of the Survey children, and then their family are. As for content to to have their class teacher contacted, and then teachers, amongst many other questions are asked compared to children of the same age. How would you rate the ability of this specific child? And it's mostly on a scale from well below average to what? Above average, apart from Scotland, that has this different

533
01:43:13.850 --> 01:43:18.810
different set up. But it still ask to the survey. Yes.

534
01:43:19.170 --> 01:43:26.139
it's not a formal teacher rating as in one that then goes into. Yeah, thank you. You're welcome

535
01:43:29.620 --> 01:43:30.990
for questions.

536
01:43:35.610 --> 01:43:36.610
I

537
01:43:39.469 --> 01:43:44.330
I have a question

538
01:43:45.340 --> 01:43:53.940
over and under evaluations may or may be explained by different magnetism. So I could imagine that

539
01:43:54.080 --> 01:44:02.330
a very high cultural capital play as a role, but only at the very top of the. Let's say that this creates

540
01:44:02.520 --> 01:44:21.999
some of the over evaluations. But for the rest of the

541
01:44:22.219 --> 01:44:52.510
another search that would make a distinction that you? Yeah, thank you. This is a great question. And it's it's definitely something that we've been thinking about at this stage. We haven't done anything in that direction, although it's true that I've said behavior behavioral issues all the way. But it's also true that some of these validated scales that we use don't necessarily just capture, for example, conduct problems, but also capture positive, non-cognitive social, emotional skills. Then so it might also, they might also have an effect.

542
01:44:52.530 --> 01:45:02.829
but still in different ways. I think this is a great point, and we definitely do have plans to explore this in different ways. So I totally get your point. Yes.

543
01:45:04.990 --> 01:45:05.790
this.

544
01:45:10.480 --> 01:45:11.200
Okay.

545
01:45:12.340 --> 01:45:15.860
to pick up the microphone on the.

546
01:45:16.090 --> 01:45:28.970
And I'd like to compliment you for the fairness presentation. my question is to do with regard to the invitation for mediation effects which you now interpret. as the reduction in in the gap between the lowest and the hires

547
01:45:29.020 --> 01:45:31.300
social class group

548
01:45:31.310 --> 01:45:50.210
the thing that I wanted. As to what extent actually did the differences from from the group to 0 actually are reduced by including these mediating effects. So in the end, after you have included all these mediators, what extent are the deviations from?

549
01:45:50.270 --> 01:46:11.199
the mean of the residual resolution? Yeah. this is also a great question. It's another thing that we're still debating along this alongside what to use as capturing ses, because we don't think this is a trivial question when you're doing cross-country comparison. So we've done some preliminary running of

550
01:46:11.200 --> 01:46:33.239
the same analysis, instead of using income quintiles using parental occupation or parental education and England and Scotland are very similar, but Germany doesn't necessarily behave in a similar way. So it's still part of what we're thinking about and what you're talking about. So in this case, for this presentation this seems like the most deceived, the most straightforward way to present what otherwise is

551
01:46:34.030 --> 01:46:48.560
a very long list of models and coefficients. But I think what you're saying. It's definitely super correct and it's something that we will. We will. We haven't done yet, but it's part of what we're planning to do. Yeah.

552
01:46:56.010 --> 01:47:12.929
thank you for your presentation. And the immense work you did on data harmonization along the way to do. I have a question that relates to the one that has been asked a minute ago that is, that you choose income for the child's socioeconomic position

553
01:47:12.930 --> 01:47:29.849
because what one I could think of, and I'm not sure if there's any literature on that is a teacher student ses match. That would add some further complication. But perhaps there is some research. I don't know if there is some research showing that it matters if it matters. I would suggest to

554
01:47:29.850 --> 01:47:34.480
have a look at it, and if it doesn't matter, you don't have to. And then you have more

555
01:47:35.450 --> 01:47:49.409
more reason to say we go for income and not for anything else. Yeah, so I can. Only I cannot necessarily say exactly what's the state of research on this? I know that from like some previous research that I've done is that

556
01:47:49.570 --> 01:48:12.600
when when you're doing some cross-country comparison at some point, income captures enough that you can compare across country sometimes in a more meaningful way the comparison of educational level. It's always a little bit hard, because it depends a bit on. Are you comparing relative distribution of education, or are you comparing, for example, an absolute level of what you consider a high or a medium medication?

557
01:48:12.600 --> 01:48:21.580
But I think this is I. I personally completely agree. Income was not our initial idea of how to capture this.

558
01:48:21.580 --> 01:48:42.019
It just makes it a lot harder when you're showing occupational categories, for example, which are a lot based on the data that we have. England and Scotland are a lot more harmonizable than together with Germany. So we're still working on this so I don't have. I. I feel like I'm answering every question which we're still working this. But it's yeah. Yeah.

559
01:48:44.210 --> 01:48:45.390
Last question

560
01:48:46.570 --> 01:49:01.189
I was wondering if there is a possibility that the child, that aptitudes and behavioral issues might be influenced by a feeling of be having biased teachers and

561
01:49:01.190 --> 01:49:19.160
and not the other way around. And I was also curious if there are any gender differences here. Because, yeah, yeah, okay, so this is a great question. on the first one in many ways. Yes, and unfortunately, the reality is that as much as we're trying to take all of the mediators from a previous time point with child attitudes.

562
01:49:19.180 --> 01:49:27.229
especially in England, but also in Scotland. We don't really have a choice. We only have it in the same wave when the teacher is waiting.

563
01:49:27.230 --> 01:49:51.059
so. Yes, it is possible. we've done some other very interesting research that I'm not presenting here on teacher bias, on attitude. And it's it seemed very clear that it's not child's attitudes that are as he has graded, but it's teacher perception of child attitudes that are as he as graded. So even here, when we're looking at the potential mediating effect

564
01:49:51.060 --> 01:49:56.639
of the attitude, it would actually be the perception of the teacher of those attitudes.

565
01:49:56.860 --> 01:50:10.779
Right? in terms of gender. Okay, so this is the overall question of this, this whole project, because we are looking at ethnic and social biases. And also we have plans to look at gender. And the question is always.

566
01:50:10.800 --> 01:50:20.770
if the focus of this specific work is on ses, do we also at least control? Or do we model interaction and intersecting effects.

567
01:50:20.860 --> 01:50:31.469
And sometimes the answer is, Yes, sometimes we decided not to do it. Now, at this stage we've decided to not do it with gender, although I think that because we're focusing on language skills, and we know that there is.

568
01:50:31.640 --> 01:50:58.900
there is a very strong gender effect. I think I will do it definitely. But overall we've tried to not just control for other things without having a framework of what and how the kind of interaction. And the modeling was happening. Because I don't know. I have a feeling that sometimes we control for things which attract them there. We don't really have substantive interpretation of how they're interacting.

569
01:51:01.820 --> 01:51:06.220
Okay, I think this ends our session. Thank you

570
01:51:06.260 --> 01:51:13.850
everyone very much, Carl, for organizing this session and all the presenters and discussion for the active contribution.

571
01:51:14.440 --> 01:51:17.639
and I would say a good night to people in Europe.

572
01:51:17.970 --> 01:51:18.630
and

573
01:51:19.300 --> 01:51:22.280
Carlo BARONE: thank you. Thank you en it.

574
01:51:22.420 --> 01:51:24.419
Carlo BARONE: Thank you. All the presenters

575
01:51:25.460 --> 01:51:26.420
Carlo BARONE: by.