On the scaling of NSW HSC marks in mathematics and encouraging higher participation in calculus-based courses

Introduction

Concern has been expressed regarding the declining number of students who take advanced calculus-based mathematics courses at the senior secondary school level. One of the reasons suggested for this development is the possibility that current scaling algorithms used to determine rankings for university admission—such as the Australian Tertiary Admission Rank (ATAR)—might disadvantage students who take some of the advanced mathematics courses. In this article, this question is explored by using data from a cohort of students in New South Wales (NSW), Australia. Data available for the students in this cohort include performance on the School Certificate (SC) Test in mathematics in Year 10 as well as their Year 12 Higher School Certificate (HSC) performance in either the HSC general mathematics or the calculus-based HSC mathematics course.

In this article, the appropriateness of the current scaling algorithm for HSC general mathematics and HSC mathematics is reviewed by comparing the differences in results obtained by students in these two groups between their SC Test in Mathematics and their scaled score obtained for their chosen HSC course in mathematics. This article begins with a discussion of some background in terms of relevant literature and information regarding the pertinent examination and scaling processes in the NSW secondary school system. Then, the research question is presented followed by a description of the data analysed and a detailed discussion of the method of analysis and results.

Background

The main tenet of this article is that students may choose easier mathematics courses to optimise the scores that are used in the university admissions process to keep open as many options for tertiary study as possible. In order to provide a context for this exploration, some related issues are discussed in this Background section. They include the benefits of studying mathematics through to the completion of secondary education and the decline in students taking mathematics and in particular higher level mathematics courses through to the final year of secondary education. To provide some background to the data that have been analysed in this article, information is also provided in this section on the Year 10 SC Test in Mathematics, the ATAR, and the approach to scaling marks from the NSW HSC examinations and moderated assessments usually occurring in Year 12 and used to produce the ATAR.

The ATAR and the approach to scaling marks from the NSW HSC examinations used to produce the ATAR

The ATAR is a single figure between 0 and 99.95 with increments of 0.05. It is a rank that is intended to measure the performance of a student in the Year 12 examinations and assessments relative to other students. If a student receives an ATAR of 80.0, it means that he or she has performed at a level better than 80% of their cohort where their cohort is the group of students who began Year 7 studies at the same time as the particular student. The calculation of the ATAR is necessarily complicated because of the wide choice of courses that students can study for their HSC examinations. Some courses will be studied by students who are a stronger group academically than other subjects. In the case of the two mathematics courses described above, one might expect that, on average, students who choose HSC general mathematics are not as strong academically as those students who choose to study the HSC mathematics course. There will, of course, be an overlap between the two groups with some students who opt for HSC general mathematics significantly stronger on a measure of overall academic ability than students in the HSC mathematics course. As argued by Hill, Brown, Rowe, and Turner (1997), in the absence of some general achievement test, it becomes necessary to scale the raw examination marks and moderated in-school assessment marks that students obtain in particular courses to allow for the strength of the competition provided by the entire cohort of students in each course. The approach used for measuring the strength of the competition in each course is to look at the performance of students in each course for all of their courses. If students who have opted to study HSC general mathematics do not perform as well as students who study HSC mathematics when compared across all courses that the students in each group study, then the HSC general mathematics raw marks and moderated school assessment marks will be scaled to a lower mean mark than for the HSC mathematics course. This is, in fact, the case as can be seen from the percentiles for scaled marks in HSC mathematics and HSC general mathematics is provided in Table 1. Further details regarding the approach to scaling of marks used to produce ATARs in NSW for the cohort in the current analysis are available at http://www.uac.edu.au/documents/atar/2013-ScalingReport.pdf.

OPEN IN VIEWER

Table 1.

Scaled course scores and overall HSC standardized scores for the maximum and various percentiles for HSC general mathematics and HSC mathematics from the Report on the Scaling of the 2013 NSW Higher School Certificate.

	Maximum	99th percentile	90th percentile	75th percentile	50th percentile	25th percentile
HSC general mathematics students
HSC general mathematics score (scaled out of 50)	45.7	41.4	35.4	29.6	21.3	13.2
Overall HSC score (standardised out of 50)	49.5	47.0	43.0	38.5	33.5	30.0
HSC mathematics students
HSC mathematics score (scaled out of 50)	50	45.8	41.4	37.9	32.7	25.8
Overall HSC score (standardised out of 50)	50	48	46	43.5	39.5	35.0

Note: HSC, Higher School Certificate; NSW, New South Wales.

Research question

Recently, a number of high school teachers, students, and academics across NSW have questioned the suitability of the scaling process used in the calculation of the ATAR which is the measure used in the admission process to Australian universities. The results of a recent survey of its members by the Mathematical Association of New South Wales (MANSW) illustrate this point. Details of the survey and a full report can be found at http://www.mansw.nsw.edu.au/resources/public-resources/2013-secondary-mathematics-teacher-survey-report. Of more than 1000 mathematics teachers surveyed, 51% believed that some students in their school were selecting senior mathematics courses below their capability. Of those surveyed, 565 teachers provided a total of 1187 reasons for students choosing a mathematics course below their capability. A desire to optimise HSC and ATAR results was the most common reason given and was cited over 200 times by these teachers.

By the way of background to the output from the scaling algorithm, we repeat some information provided in the Report on the Scaling of the 2013 NSW Higher School Certificate (see http://www.uac.edu.au/documents/atar/2013-ScalingReport.pdf) in Table 1. The information in Table 1 shows the scaled mark (out of 50) and underneath the HSC mark (also out of 50) obtained by students at the top, at the 99th percentile (P99), at the 90th percentile (P90), at the 75th percentile (P75), at the 50th percentile (P50), and at the 25th percentile (P25) in each of the two courses HSC general mathematics and HSC mathematics.

From Table 1, it can be observed that a student performing at the 90th percentile of HSC general mathematics is awarded a scaled score (of 35.4) approximately equal to a student in HSC mathematics performing at the 60th percentile (between 32.7 and 37.9) or perhaps slightly better than this. Given that teachers and students have historically regarded HSC general mathematics as the option for weaker students, this may appear to give those weaker students an unfair advantage relative to those who take the HSC mathematics course. Of course, it is possible that students who perform well at mathematics are opting to take the easier HSC general mathematics in order to free up time for other courses which interest them more and the above scaled marks may accurately reflect the relative performance of the two groups of students. We investigate this question in this article.

It is, therefore, of interest to compare the marks obtained by students in the Year 10 SC Test in Mathematics with their estimated scaled marks obtained at the completion of their Year 12 studies. We examine whether the relationship between these two numerical measures of performance is the same or different for students who choose HSC general mathematics compared to those who choose HSC mathematics.

Prior to reporting on the results of this analysis, we note that the 90th percentile of students enrolled in HSC general mathematics was an aligned score of 83 out of 100 on the Year 10 SC Mathematics Test. The 60th percentile of students enrolled in HSC mathematics was an aligned score of 88 out of 100 on the Year 10 SC Mathematics Test. This difference of very close to half one standard deviation (SD) in Year 10 SC Mathematics Test scores (see Table 2) is significant given that these two scores are subsequently aligned very closely by the HSC scaling algorithm. Looking at this a different way, the 60th percentile of marks on the SC Mathematics test for students who take HSC Mathematics is equal to the 98th percentile of marks on that same test for students who take HSC general mathematics. Conversely, the 90th percentile of marks on the Year 10 SC Mathematics test for students who take HSC general mathematics is at the 30th percentile of marks on that same test for students who take HSC mathematics. These observations provide a strong motivation for the analysis to examine the appropriateness of the current scaling algorithm that is provided in the following two sections. We seek to determine whether the relationship between Year 10 achievement in mathematics and the scaled score obtained on mathematics in Year 12 differs depending on which level of mathematics is taken at the HSC level.

OPEN IN VIEWER

Table 2.

Summary statistics for cohort data.

	Number of students	Year 10 Maths SC (aligned out of 100) test mean	Year 10 Maths SC (aligned out of 100) test SD	Year 12 standardized HSC (out of 50) mean	Year 12 HSC standardized (out of 50) SD
HSC general mathematics	30,631 (68.1%)	70.8	8.64	33.9	6.91
HSC mathematics	14,335 (31.9%)	85.6	7.14	38.7	6.03
Both HSC courses	44,966	75.5	10.71	–	–

Data

The data used in the study reported here were taken from the BOSTES. The data included the SC Mathematics test results, generally taken in Year 10, as an aligned mark out of 100, for every student who completed this test during the years 2009, 2010, and 2011. Raw marks on this test had been aligned to performance bands with a mean score of about 75. In addition to these data, for every student with a SC result over this period, and who subsequently was examined in either HSC general mathematics or HSC mathematics over the period up to and including 2013, the student's standardised HSC mark was obtained. The standardised HSC mark was derived as an equally weighted combination of the moderated school assessment mark and the final examination mark. Some students who completed the Year 10 SC test did not have a mark recorded for their HSC mathematics in these data, either because they did not complete any mathematics during Year 12, or because they completed a different mathematics course in Year 12 or did not complete Year 12 studies during the follow-up period.

The analysis that is described in the following section uses data from the SC in 2011 and from the HSC in 2013. The analysis has been repeated with other years of data and the results presented below, under Results and discussion are repeated without significant differences and are therefore not reported.

In total, 83,184 students sat the Year 10 SC Mathematics Test in 2011. These students were tracked through until the end of calendar year 2013. Of these students, 20,687 did not complete the HSC during the follow-up period and did not study either HSC mathematics or HSC general mathematics by the end of 2013. A further 11,518 of these students did complete the requirements for award of the HSC by the end of 2013 but did not include any mathematics in their HSC-level studies. Other mathematics courses (either mathematics extension 2, mathematics applied, or mathematics life skills) were completed by 5691 of these students by the end of 2013. Finally, 412 of these students completed an HSC course in mathematics during 2012 rather than 2013. Removing these students from the 83,184 students who sat the SC Test in Mathematics in 2011, left 44,966 students who also completed either HSC general mathematics or HSC mathematics in 2013. It is the marks from this group of students who form the data set used in the analysis reported in this article.

Table 2 provides summary information on the number of students from our Year 10 cohort taking HSC general mathematics and HSC mathematics during 2013. The mean and SD of the marks obtained by these students in both the SC Mathematics test (as an aligned mark) and their Year 12 mathematics course (as a standardised mark out of 50) are also given. Aligned marks and standardised marks are produced in similar ways by relating the performance on the relevant examination to a set of qualitative performance bands. The terminology used is consistent with what has been used in NSW over the period for which data have been collected. Information for the two groups of HSC students combined is also given where appropriate.

From the above summary statistics, we see that just over two thirds of our cohort studied HSC general mathematics. The performance of the students who studied HSC general mathematics was substantially lower than those who studied HSC mathematics on the SC Test in mathematics. The HSC summary statistics reported in Table 2 represent standardised marks.

In Figure 1, a histogram, shows the relative frequency (or “density”) of marks obtained by students who studied HSC general mathematics on their SC Mathematics Test. Overlaying this histogram are two estimated density functions. The solid line shows the estimated density function for marks obtained by HSC general mathematics students on the SC Mathematics Test. We observe that this estimated density function tracks close to the histogram. We also plot an estimated density function of marks obtained by students who studied HSC mathematics on their SC Mathematics Test using a broken line. It is clear from Figure 1 that, as we would expect, HSC mathematics was chosen by students who did much better overall on the SC Mathematics Test than students who chose HSC general mathematics. There is, of course, some overlap between the SC mathematics performance of the two groups of students. The top performing students who opted for general mathematics at HSC level outperformed those students who chose HSC mathematics and performed at the lower end of this group on the SC Mathematics Test. OPEN IN VIEWER

Figures 2 and 3 provide scatterplots of overall HSC results against SC Test results in HSC general mathematics and HSC mathematics. Areas on the scatterplots with a greater number of points are shown as darker than areas where fewer points are located. The positive correlation between SC performance and HSC results is clear for both the HSC general mathematics group and the HSC mathematics group of students. OPEN IN VIEWER

OPEN IN VIEWER

Figure 3.

Scatterplot of HSC and SC marks for students who studied HSC mathematics.

Figures 2 and 3 show a clear strong positive correlation between the score obtained by students on the Year 10 SC Mathematics Test and their subsequent HSC performance as measured by their standardised—but not scaled—score at Year 12. Some students clearly perform better at the SC level relative to their HSC score and some perform less well. The SC test can, therefore, be considered not a perfect but still a reliable predictor of performance at the HSC level. Variation in the points around the lines of best fit provided in Figures 2 and 3 occurred due to other factors that could have been included into our model such as, for example, students working harder in Years 11 and 12 than in earlier years of education or students developing and maturing earlier or later than other students. Of course, some of the variation about the lines is due to random variation and the imperfect nature of the SC test and HSC exams as measures of mathematical achievement.

Method

In order to compare the different relationships between the scaled HSC mark obtained in mathematics and the Year 10 SC test mark for students who study HSC general mathematics and those who study HSC mathematics, we first needed to produce estimated scaled scores for students in the two HSC mathematics courses.

The BOSTES data set provided the HSC mark obtained on the two courses but not the scaled mark used to derive the ATAR due to privacy reasons. Similarly, it was not possible to obtain a table converting HSC marks to scaled marks for each course—again because of privacy rules. In 1998, the NSW Government passed legislation that prevented the reporting of Universities Admission Indices—now ATARs—to schools, the Board of Studies, or the media. More recently, on application, students have been able to obtain the raw subject marks on HSC examinations. If students had both raw marks and a table converting these to scaled marks they could, theoretically, recreate aggregates of scaled scores that are used to derive ATARs. This would go against the legislation passed by the NSW Government in 1998, hence the need for privacy. Therefore, we estimated the scaled scores using publicly available information, specifically the information provided in Table 1.

In order to estimate scaled scores, we used linear interpolation to form a relationship between HSC standardised marks and scaled scores using the percentile information given in Table 1. Then, we calculated the mean and SD of the estimated scaled marks produced from the linear interpolation process. These were very close to the published mean and SD of the actual scaled marks used to determine the ATARs, which provided confidence in the estimation approach.

It should be noted that the estimation here did introduce some imperfection in the data used for the analysis—specifically since the actual scaling process worked directly with raw examination and moderated assessment scores to produce scaled marks instead of working from HSC marks produced by the BOSTES. Different—adjacent—raw marks may map to a single HSC mark, which means that these two approaches to scaling will differ. In courses with large enrolments and a wide spread of HSC marks, this will be less problematic than it may first appear. In order to determine the extent of imperfection introduced into the analysis by this approach, an email discussion with the Chair of the Technical Committee on Scaling for the NSW HSC was held (Weber, 2014). The communication concluded that the linear interpolation approach to producing scaled scores above the lower quartile should work very well for courses with large enrolments like HSC general mathematics and HSC mathematics. Further, it was concluded that scaled marks awarded in the lowest performance band for each course would deviate somewhat from the linear assumption used in this article. However, with so few marks in this region, the effect on summary statistics such as means obtained from scaled score estimates was considered to be negligible. It is also worth noting that what is most significant here is that the relationship between the true scaled scores and the estimated scaled scores is similar across the two HSC courses. Given the assurance that the approach used works very well and the close alignment of the mean and SDs of the estimated and actual scaled scores, we can be confident in the analysis presented in this article. Importantly, given the strong evidence for the conclusions that can be drawn from the analysis which follows, we are very confident that the key messages coming from this work would not change as we had used the actual scaled marks which were not available.

Results and discussion

Figure 4 shows a histogram of the estimated scaled marks for students taking HSC general mathematics along with an estimate of the probability density functions for the scaled marks for students in both HSC general mathematics and HSC mathematics. First note that the estimated density functions are positive at values where there are no values in the data. For example, we know that the maximum scaled score for HSC general mathematics is 45.7 and yet there is some positive density in the region to the right of 45.7. Similarly, we observe some positive estimated density for scaled scores less than zero. This is a function of the simple density estimation method used here. The purpose of the density estimation is to give a graphical view of the location and relative frequencies of scaled marks. More precise statistical tests are applied later in the analysis. OPEN IN VIEWER

At first glance, Figure 4 looks similar to Figure 1. Upon closer inspection, however, an important difference emerges. In Figure 1, the mode of the estimated density function for the SC mathematics students' scaled scores for students who studied HSC mathematics occurs where the density estimate for the HSC general mathematics students is about 0.01. In Figure 4, the corresponding mode occurs where the density for HSC general mathematics students is about 0.02. There is considerably more overlap in the regions of relatively high density in Figure 4 than in Figure 1. This provides some graphical evidence that students in HSC mathematics are awarded scaled scores which are lower relative to their Year 10 SC test result than for students who take HSC general mathematics.

Given that the SC Mathematics Test results are reported out of 100 and that HSC marks are reported out of 50 and that each has different coefficients of variation, we standardised the scores on each test by subtracting the mean test result from each score and dividing the resulting difference by the SD of the test scores. This gives a standardised score or z-score whereby a score of “0” denotes a score equal to the average score obtained by all students who took that particular test or HSC course. A z-score of “1” indicates a score obtained which is exactly one SD above the mean. Similarly, a z-score of −2 indicates a very low score equal to the mean minus 2 SDs. For each student, two z-scores are produced. The first z-score relates to their Year 10 SC Mathematics test result and the second z-score relates to their Year 12 HSC estimated scaled score on the relevant mathematics course.

We examine the z-scores obtained as described in the previous paragraph in a number of ways. First the proportion of students in each HSC course who saw an improvement in their z-score from their SC test to their HSC scaled score is calculated. We calculate the proportion of students who have a substantial improvement (defined as an increase of greater than 0.5) in their z-score and similarly the proportion of students who have a substantial reduction in their z-score from the SC test to the HSC estimated scaled result for each completed HSC course. Second, we perform a chi-squared test to determine if the expected distribution of the z-scores obtained in the HSC—where expected values are obtained using the performance in the Year 10 SC Mathematics test—differ significantly from the distribution of z-scores actually obtained by students in each HSC course. Finally, we report a regression analysis with the z-score obtained on the HSC course regressed on the SC Mathematics Test z-score obtained and a separate indicator function for whether or not the student studied HSC general mathematics. This sort of regression analysis, with both a continuous dependent variable and an indicator variable as predictors is a form of analysis of covariance.

Tables 3 and 4 show the results of the first two analyses of z-scores. From Table 3, it is clear that the relative performance of students, as measured by the standardised estimated HSC scaled score and standardised SC test result improves about twice as often for students who take HSC general mathematics than for those who take HSC mathematics. It can also be seen in Table 3 that while big improvements are rare for both courses, a full one quarter of all students who take HSC general mathematics see an improvement in their z-score of greater than half of a SD, while the HSC mathematics students saw such an improvement only 7.3% of the time. The observation for students who see a big reduction in their z-scores is similar with HSC general mathematics far less likely to have a large reduction in their z-score than students who study HSC mathematics.

OPEN IN VIEWER

Table 3.

Analysis of change in standardized scores from the SC test to the HSC scaled results.

	Proportion who improve z-score (%)	Proportion who improve z-score by more than 0.5 of a standard deviation (%)	Proportion who have z-score reduce by more than 0.5 of a standard deviation (%)
HSC general mathematics	59.7	25.0	11.1
HSC mathematics	29.6	7.3	34.6

OPEN IN VIEWER

Table 4.

Frequencies of z-scores for HSC general mathematics and HSC mathematics and cumulative percentages.

	HSC general mathematics, SC Test z-scores (expected)	HSC mathematics, SC Test z-scores (expected)	HSC general mathematics scaled, z-scores (observed)	HSC mathematics scaled, z-scores (observed)
<−2.5	26 (0.1%)	1 (0.0%)	0 (0.0%)	0 (0.0%)
[−2.5, −2)	499 (1.7%)	1 (0.0%)	994 (3.2%)	128 (0.9%)
[−2, −1.5)	2752 (10.7%)	36 (0.3%)	1538 (8.3%)	69 (1.4%)
[−1.5, −1)	4746 (26.2%)	76 (0.8%)	5885 (27.5%)	147 (2.4%)
[−1, −0.5)	7130 (49.5%)	367 (3.4%)	5067 (44.0%)	805 (8.0%)
[−0.5, 0)	5812 (68.4%)	784 (8.8%)	5534 (62.1%)	1872 (21.1%)
[0, 0.5)	5295 (85.7%)	1861 (21.8%)	4177 (75.7%)	2716 (40.0%)
[0.5, 1)	3443 (97.0%)	4158 (50.8%)	4191 (89.4%)	3332 (63.3%)
[1, 1.5)	818 (99.6%)	4103 (79.4%)	2657 (98.1%)	3572 (88.2%)
[1.5, 2)	106 (100.0%)	2416 (96.3%)	588 (100.0%)	1595 (99.3%)
[2, 2.5)	4 (100.0%)	532 (100.0%)	0 (100.0%)	99 (100.0%)
≥2.5	0 (100.0%)	0 (100.0%)	0 (100.0%)	0 (100.0%)

Note: SC, School Certificate; HSC, Higher School Certificate.

Table 4 shows the number of students with standardized z-scores in bands <−2.5, [−2.5, −2.0), …, [2.0, 2.5), and >=2.5. There are four groupings, shown as the four columns in the table, namely the results on the SC Mathematics test for each HSC course taken and the scaled results on each HSC course taken. The percentage figures, quoted in brackets, are the cumulative proportion of students with a z-score at the level shown or lower. Considering first the HSC general mathematics results, we see that 68.4% of students in this group scored below average—where the average is calculated across the students who took the HSC general mathematics or HSC mathematics examination two years later—on the SC Mathematics test. Only 62.1% of these students were awarded an estimated scaled score, which was below the average of the group. Looking to the HSC mathematics results, we see that only 8.8% of these students scored below average on the Year 10 SC Mathematics test, while 21.1% of them were awarded a scaled mark lower than the average of these students at the HSC level.

Suppose we assume that the performance on the Year 10 SC test can be used to predict the performance of students in the HSC Mathematics exams. If this applies, the expected number of students with scaled HSC z-scores in each of the 12 bands in Table 4 can be set as the observed numbers in each band based on the Year 10 SC test (see first two columns in Table 4). This expected distribution of the z-scores can then be compared with the observed estimated distribution of z-scores actually awarded to students after scaling at the HSC level using a chi-squared test. Results show a significant difference in the two distributions at the 0.1% significance level with a test statistic well in excess of 1000 for the two analyses considered. The test statistic is calculated as the usual sum of squared differences between observed and expected cell counts divided by the expected cell count in each cell. The test is applied with 11 degrees of freedom—the usual application of this test where degrees of freedom are equal to the number of groups, here 12—as can be seen in Table 4—minus 1. Those students who take HSC general mathematics have observed z-scores for their HSC performance significantly higher than we would expect. Those students who take HSC mathematics have observed z-scores for their HSC performance significantly lower than we would expect.

One of the criticisms of the chi-squared test, as applied above, is that the data are arbitrarily grouped prior to analysis—here into bands of width one half SD each—resulting in some loss of precision.

An alternative method of analysis, motivated by data provided in Figures 2 and 3, is to perform an analysis of covariance or regression analysis. We use the HSC scaled z-score as the dependent variable in the regression given in equation (1). As can be seen in equation (1), an intercept, denoted β 0 , a linear term being the z-score obtained on the SC Mathematics test, denoted coefficient β 1 and an indicator variable, denoted “general,” which is 1 if the student studied HSC general mathematics and 0 if the student studied HSC mathematics. This general variable is multiplied by a coefficient β 2 in the model equation.

E [ z -score for HSC course in ] = β 0 + β 1 ( z -score for SC Mathematics Test ) + β 2 ( General )

(1)

The coefficients β 0 , β 1 , and β 2 are estimated using ordinary least squares regression. The estimated regression coefficients along with their standard errors and p-values for a test of significance of each of these coefficients being different from zero are given in Table 5.

OPEN IN VIEWER

Table 5.

Analysis of covariance results.

Coefficient	Coefficient value	Standard error	p-value
Intercept ( β 0 )	−0.2398	0.00571	<0.001
SC Mathematics Test ( β 1 )	0.9216	0.00347	<0.001
General ( β 2 )	0.3513	0.00743	<0.001

The results in Table 5 show that the estimated coefficient of the slope of the SC Mathematics Test is positive and statistically significant (0.9216). This is to be expected as a higher SC Mathematics Test result is associated with a higher HSC mathematics scaled z-score regardless of whether the student takes HSC general mathematics or HSC mathematics. In addition, the general indicator variable has a positive estimated coefficient of 0.3513. This result indicates a positive effect on the dependent variable purely from having studied the HSC general mathematics course, after taking into account the effect of the SC Mathematics Test z-score on the estimated HSC scaled z-score. This positive effect is statistically significant and also of great practical significance. An estimated increase of 0.3513 SDs is important from a practical perspective and has the potential to influence students' decisions regarding which mathematics course to choose, given that their decision to enrol in HSC mathematics or the HSC general mathematics course can be shown to have an impact on their Year 12 outcome.

One other way to illustrate the results graphically is to calculate the change in z-scores from the Year 10 SC test to the HSC scaled mark for students in both HSC courses. We compare these two sets of data on change in z-score using summary statistics in addition to estimated density functions for each set of data. While Table 6 shows the summary statistics, Figure 5 shows the estimated density functions for the change in z-scores. OPEN IN VIEWER

OPEN IN VIEWER

Table 6.

Summary statistics for change in z-score by course.

	Mean change in z-score	Standard deviation of change in z-score
HSC general mathematics	0.148	0.552
HSC mathematics	−0.316	0.600

Note: HSC, Higher School Certificate.

As can be seen in Table 6, the average change in z-score obtained when comparing the SC Mathematics test to the HSC scaled mark is positive for HSC general mathematics and quite large and negative for HSC mathematics. It is clear from Figure 5 that students who take HSC general mathematics are considerably more likely to receive a higher increase (or smaller decrease) in their z-score compared to students who take the HSC mathematics course.

The analysis presented here shows that the increase in performance—as measured by the change in standardised score from the SC Test to the HSC scaled mark—is considerably higher for students who opt to take HSC general mathematics than for those who take HSC mathematics. The results indicate that for many students, particularly the lower performers in the HSC mathematics examination, it may well be better, to take HSC general mathematics instead of HSC mathematics, if their aim was to maximise their scaled mark for mathematics, which contributes directly to their ATAR.

Conclusion

This article provides an analysis of the relative performance of students in the Year 10 SC Test in mathematics during 2011 and their estimated scaled score in their chosen HSC mathematics course for students who sat the HSC general mathematics examination or the HSC mathematics examination in 2013. The results indicate that, on average, the study of HSC general mathematics leads to materially higher scaled scores, relative to the performance on the SC Mathematics test, than does the study of HSC mathematics. While not all students stand to benefit from choosing HSC general mathematics over HSC mathematics, the current scaling mechanism provides a strong incentive to take HSC general mathematics for a very large group of students. At a time when many are deeply concerned about the reducing numbers of students studying higher level mathematics in the final year of secondary education, it is useful to consider the evidence presented here which supports one of the possible explanations for this drop in numbers. It is hoped that this article will contribute to the debate on the appropriateness of the current scaling algorithm in respect of HSC general mathematics and HSC mathematics. In addition, it seeks to contribute to the wider discussion around higher level mathematics courses at senior secondary school, both in terms of students' uptake and the potential need for universities to (re-)introduce prerequisites to prevent students from struggling with the demand of certain courses.