SUPPORT DOCUMENT
Adjustment notes for
apprentice and trainee
estimates: March
quarter 2015
National Centre for Vocational Education Research
This technical note provides a cumulative record of the
adjustment notes relevant for the estimation of apprentice and
trainee figures at each collection. This document commences with
the adjustment note for Collection 84 (June 2015 estimates used
to produce the publication, Australian vocational education and
training statistics: apprentices and trainees 2015 — March
quarter, available at <http://www.ncver.edu.au>).
The views and opinions expressed in this document are those of NCVER and do
not necessarily reflect the views of the Australian Government, or state and
territory governments.
© Commonwealth of Australia, 2015
With the exception of the Commonwealth Coat of Arms, the Department’s logo, any material
protected by a trade mark and where otherwise noted all material presented in this document is
provided under a Creative Commons Attribution 3.0 Australia
<http://creativecommons.org/licenses/by/3.0/au> licence.
The details of the relevant licence conditions are available on the Creative Commons website
(accessible using the links provided) as is the full legal code for the CC BY 3.0 AU licence
<http://creativecommons.org/licenses/by/3.0/legalcode>.
The Creative Commons licence conditions do not apply to all logos, graphic design, artwork and
photographs. Requests and enquiries concerning other reproduction and rights should be directed to
the National Centre for Vocational Education Research (NCVER).
This document should be attributed as NCVER 2015, Adjustment notes for apprentice and trainee
estimates: March quarter 2015, NCVER, Adelaide.
This work has been produced by NCVER on behalf of the Australian Government, and state and
territory governments, with funding provided through the Australian Department of Education and
Training.
The views and opinions expressed in this document are those of NCVER and do not necessarily reflect
the views of the Australian Government or state and territory governments.
Published by NCVER, ABN 87 007 967 311
Level 11, 33 King William Street, Adelaide, SA 5000
PO Box 8288 Station Arcade, Adelaide SA 5000, Australia
Phone +61 8 8230 8400
Fax +61 8 8212 3436
Email ats_req@ncver.edu.au
Follow us:
Web <http://www.ncver.edu.au> <http://www.lsay.edu.au>
<http://twitter.com/ncver>
<http://www.linkedin.com/company/ncver>
Contents
2
Introduction
4
Adjustment note for Collection 84
5
Attachment 1: Revised estimates for Collection 84
8
Attachment 2: Expired contracts for Collection 84
11
Adjustment note for Collection 83
14
Attachment 1: Revised estimates for Collection 83
16
Attachment 2: Expired contracts for Collection 83
17
Adjustment note for Collection 82
22
Attachment 1: Revised estimates for Collection 82
24
Attachment 2: Expired contracts for Collection 82
26
Adjustment note for Collection 81
29
Attachment 1: Revised estimates for Collection 81
31
Attachment 2: Expired contracts for Collection 81
33
National Centre for Vocational Education Research
3
Introduction
Apprentice and trainee data are reported by the State and Territory Training Authorities to NCVER on
a quarterly basis, starting at the September quarter of 1994. The set of data submitted that quarter is
referred to as Collection 1. The sets of data submitted in subsequent quarters are referred to as
Collection 2, Collection 3 and so on.
NCVER publishes data on the numbers of contracts of training that commence, complete,
cancel/withdraw, re-commence, expire or are suspended and the time at which these events occur
(referred to as the "date of effect"). From these events, the number of contracts in training at a given
time can be calculated.
Due to time delays in reporting data on the status of contracts to NCVER, the most recent data are
estimated. In short, the estimation methodology is based on the calculation of “average lag ratios”. A
lag ratio is the ratio of the actual number of events (commencements, completions, etc) which
occurred in a particular quarter to the number of those events which were reported in a given
quarter. The average lag ratio is calculated by taking the average of the lag ratios found in a “time
window”, which is a moving period of eight quarters from the past. Further details on this
methodology are provided in the technical paper produced by NCVER, Estimation of apprentice and
trainee statistics, which may be found on the NCVER Portal as a related item to this quarterly
publication.
The purpose of this technical paper is to document the adjustments that are made to the estimates at
each collection, and produce a cumulative document of these adjustments, commencing at Collection
81, September 2014 estimates.
4
Adjustment notes – March quarter 2015
Adjustment note for Collection 84
NCVER examines the quarterly apprentice and trainee estimates produced by the endorsed model in
order to check that the estimates are reasonable. In particular, a decision rule was introduced in
Collection 45 that mandated reviewing all estimates with relative prediction errors of 10% or more.
The goal of the review is to correct for any large bias in estimation that might be caused by changes
in the pattern of reporting practices over time. Note that whilst an estimate might be adjusted for
bias, its associated prediction error is not altered.
The relative prediction errors varied across the jurisdictions with some maintaining low variability
while others showing higher variability than usual. The increases were mostly moderate but there
were also some that were quite large.
The estimates for South Australia need to be treated with caution. The high relative prediction errors
noted below are thought to be a combination of factors. The time windows used to calculated average
reporting lags correspond to the transition to a new processing system and also rapid changes in nontrade commencements. Indications are that current reporting lags are lower than the averages
calculated from the time windows. This suggests that in addition to the high variability, the estimated
commencements, completions and cancellations/withdrawals could have a bias toward overestimating
the actual counts. Eventually the time window will move out of this period but estimates will be
closely monitored for some quarters yet.

The commencement estimate for South Australia for the March quarter 2015 was associated
with a relative prediction error of 12.6%.

The completion estimate for South Australia for the March quarter 2015 was associated with a
relative prediction error of 24.6%.

The cancellation/withdrawal estimate for the South Australia for the March quarter 2015 was
associated with a relative prediction error of 18.9%.

The cancellation/withdrawal estimate for the South Australia for the December quarter 2014
was associated with a relative prediction error of 24.0%.

The cancellation/withdrawal estimate for the South Australia for the September quarter 2014
was associated with a relative prediction error of 10.2%.

The cancellation/withdrawal estimate for the Northern Territory for the March quarter 2015
was associated with a relative prediction error of 24.1%.
The relative prediction errors for expired contracts remained relatively high. Tasmania (33.3%), the
Australian Capital Territory (22.2%), Western Australia (16.9%) and Queensland (15.2%) had the
highest prediction errors, while the remaining states were below 10%.
The contribution of expired contracts to the in-training estimate is usually small both in level and
variation. High relative errors appear to be explained to some degree by the fact that the estimates
are small numbers and therefore any variation is relatively large. Adjustments to the estimates of
expired contracts have little effect on the corresponding estimates of in-training. Consequently, no
alterations to estimates of expired contracts have been made.
National Centre for Vocational Education Research
5
South Australia
Commencements for the March quarter 2015
From endorsed model – Estimate = 3911; Relative error = 12.6%.
Time window for calculating the average lag factor is from June quarter 2012 to March quarter 2014.
The first three lag ratios start low and rise to a higher level at quarters five and six after which the
lag ratios decline again. Looking at the next quarters to come into the time window can be expected
to increase over the next two quarters so it is not clear how to adjust the lag ratios to improve the
estimate.
No revision.
Completions for the December quarter 2015
From endorsed model – Estimate = 2985; Relative error = 24.6%.
Time window for calculating the average lag factor is from June quarter 2012 to March quarter 2014.
The lag ratio corresponding to quarter five is a clear outlier and has been excluded in the adjusted
estimate. The lag ratio for quarter six is a possible candidate for exclusion but has been retained in
keeping with a cautious approach to adjusting estimates.
Revised estimate = 2767.
Cancellations/withdrawals for the March quarter 2015
From endorsed model – Estimate = 1469; Relative error = 18.9%.
Time window for calculating the average lag factor is from June quarter 2011 to March quarter 2013.
The first seven lag ratios display only slight variation but the eighth lag ratio is clearly higher than the
rest. However, looking at quarters nine and ten, which will come into the time window over the next
two quarters, implies that lag ratios will fluctuate over the coming quarters. It is not clear from the
data how to adjust the estimates.
No revision.
Cancellations/withdrawals for the December quarter 2014
From endorsed model – Estimate = 1362; Relative error = 24.0%.
Time window for calculating the average lag factor is from June quarter 2011 to March quarter 2013.
The first six lag ratios are relatively flat but rise in quarter seven and reach a high point in quarter
eight. Looking at the next two quarters to come into the time window suggests that the lag ratios will
remain high, at least in the short term. It is not clear yet how high this level will be so it is not clear
how to best adjust the estimate.
No revision.
6
Adjustment notes – March quarter 2015
Cancellations/withdrawals for the September quarter 2014
From endorsed model – Estimate = 1271; Relative error = 10.2%.
Time window for calculating the average lag factor is from June quarter 2011 to March quarter 2013.
The lag ratios corresponding to quarters seven and eight are at a higher level than the ratios for the
other six quarters. Looking at the next two quarters to come into the time window suggests that the
lag ratios will remain high, at least in the short term. As the time window advances over the next two
quarters a better judgement can be made as to whether the lag ratios remain high or return to a
lower level.
No revision.
Northern Territory
Cancellations/withdrawals for the March quarter 2015
From endorsed model – Estimate = 519; Relative error = 24.1%.
Time window for calculating the average lag factor is from June quarter 2011 to March quarter 2013.
The lag ratio for quarter one is a clear outlier, much higher than the others. There is a clear
downward trend through the time window but the next two quarters to come into the time window
suggest that the ratios could rise again perhaps up to the quarter two value. The lag ratio for quarter
one has been excluded in the adjusted estimate.
Revised estimate = 481.
National Centre for Vocational Education Research
7
Attachment 1: Revised estimates for Collection 84
The following graphs depict the pattern of the lag ratios for the estimates that were revised or
considered for revision. The graph shows the lag ratios for the eight quarters in the time window used
in the endorsed model (labelled 1 to 8) and also the two quarters following (labelled 9 and 10).
Horizontal lines are also displayed on the graphs. One represents the average lag as calculated from
the lags in the time window (purple line). Where there is another, it represents the average lag as
calculated from the alternative time period used for the revised estimate (black line).
South Australia commencements March 2015
1.4
1.2
Lag ratios
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
6
7
8
9
10
8
9
10
Quarters
Lag ratios
Average ratio
South Australia completions March 2015
Lag ratios
2.0
1.5
1.0
0.5
0.0
1
2
3
4
5
6
7
Quarters
Lag ratios
8
Average ratio
Alternative average
Adjustment notes – March quarter 2015
South Australia cancellations/withdrawals March 2015
5.0
Lag ratios
4.0
3.0
2.0
1.0
0.0
1
2
3
4
5
6
7
8
9
10
Quarters
Lag ratios
Average ratio
South Australia cancellations/withdrawals December 2014
2.0
Lag ratios
1.5
1.0
0.5
0.0
1
2
3
4
5
6
7
8
9
10
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
9
South Australia cancellations/withdrawals September 2014
1.4
1.2
Lag ratios
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
6
7
8
9
10
Quarters
Lag ratios
Average ratio
Northern Territory cancellations/withdrawals March 2015
1.8
1.6
1.4
Lag ratios
1.2
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
6
7
8
9
10
Quarters
k=1
10
average
Alt. ave.
Adjustment notes – March quarter 2015
Attachment 2: Expired contracts for Collection 84
Although subject to high relative errors, estimates of expired contracts have not been altered because
they are such a small contributor to the in-training estimate. As can be seen from the following
graphs, which depict the pattern of the lag ratios for the estimates of expired contracts, an
alternative way of estimating expired contracts is often unclear.
The graphs show the lag ratios for the eight quarters in the time window used in the endorsed model
(labelled 1 to 8). A horizontal line is also displayed, representing the average lag as calculated from
the lags in the time window (purple line).
New South Wales
0.4
0.3
Lag ratios
0.3
0.2
0.2
0.1
0.1
0.0
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Victoria
0.4
0.35
Lag ratios
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
11
Queensland
0.5
0.4
0.4
Lag ratios
0.3
0.3
0.2
0.2
0.1
0.1
0.0
1
2
3
4
5
6
7
8
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
South Australia
0.5
0.5
0.5
0.4
Lag ratios
0.4
0.4
0.4
0.4
0.4
0.4
0.4
0.4
1
2
3
4
5
Quarters
Lag ratios
Average ratio
Western Australia
0.3
0.3
Lag ratios
0.2
0.2
0.1
0.1
0.0
1
2
3
4
5
Quarters
Lag ratios
12
Average ratio
Adjustment notes – March quarter 2015
Tasmania
0.4
0.4
Lag ratios
0.3
0.3
0.2
0.2
0.1
0.1
0.0
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Australian Capital Territory
1.4
1.2
Lag ratios
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
13
Adjustment note for Collection 83
NCVER examines the quarterly apprentice and trainee estimates produced by the endorsed model in
order to check that the estimates are reasonable. In particular, a decision rule was introduced in
Collection 45 that mandated reviewing all estimates with relative prediction errors of 10% or more.
The goal of the review is to correct for any large bias in estimation that might be caused by changes
in the pattern of reporting practices over time. Note that whilst an estimate might be adjusted for
bias, its associated prediction error is not altered.
The commencement estimate for South Australia for the December quarter 2014 was associated with
a relative prediction error of 13.9%.
The completion estimate for South Australia for the December quarter 2014 was associated with a
relative prediction error of 26.1%.
The cancellation/withdrawal estimate for the Northern Territory for the December quarter 2014 was
associated with a relative prediction error of 23.0%.
The commencement estimate for the Australian Capital Territory for the December quarter 2014 was
associated with a relative prediction error of 8.2%.
The relative prediction errors for expired contracts remained relatively high. Tasmania (34.4%), the
Australian Capital Territory (23.2%), Queensland (18.0%), and Western Australia (17.0%) had the
highest prediction errors, while the other states and territories were below 10%.
The contribution of expired contracts to the in-training estimate is usually small both in level and
variation. High relative errors appear to be explained to some degree by the fact that the estimates
are small numbers and therefore any variation is relatively large. Adjustments to the estimates of
expired contracts have little effect on the corresponding estimates of in-training. Consequently, no
alterations to estimates of expired contracts have been made.
South Australia
Commencements for the December quarter 2014
From endorsed model – Estimate = 2760; Relative error = 13.9%.
Time window for calculating the average lag factor is from March quarter 2012 to December quarter
2013.
The lag ratios corresponding to quarters six and seven are at a higher level than the ratios for the
other six quarters. However, the lag ratios for quarters 8, 9, and 10 do not confirm that the high level
shown in quarters 6 and 7 will persist. There is also doubt as to what level the lag ratios will fall to as
the time window advances, especially as the lag ratios for quarters 9 and 10 are not considered final.
The data presents no clear basis for adjusting the estimate.
No revision.
14
Adjustment notes – March quarter 2015
Completions for the December quarter 2014
From endorsed model – Estimate = 2907; Relative error = 26.1%.
Time window for calculating the average lag factor is from March quarter 2012 to December quarter
2013.
The revised estimate was calculated by excluding the lag ratio corresponding to quarter six of the
time window (see attachment 1). The lag ratio for this quarter is clearly higher than all the other lag
ratios in the time window. Additionally, the lag ratios for quarters none and ten, though not yet
‘finalised’, strongly suggest that lag ratios will remain lower than the excluded quarter’s ratio, and
that an eventual return to the level of quarters one to five is possible.
Revised estimate = 2683.
Northern Territory
Cancellations/withdrawals for the December quarter 2014
From endorsed model – Estimate = 505; Relative error = 23.0%.
Time window for calculating the average lag factor is from March quarter 2011 to December quarter
2012
The revised estimate was calculated by excluding the lag ratio corresponding to quarter two of the
time window (see attachment 1). The lag ratio for this quarter is clearly higher than all the other lag
ratios in the time window. Although the lag ratios for quarters nine and ten are not yet considered
‘final’, the strongly suggest that lag ratios will remain lower than the excluded quarter’s ratio, and
that ratios might persist at the level of quarters four to eight.
Revised estimate = 472.
Australian Capital Territory
Commencements for the December quarter 2014
From endorsed model – Estimate = 608; Relative error = 8.2%.
Time window for calculating the average lag factor is from March quarter 2012 to December quarter
2013.
The revised estimate was calculated by excluding the lag ratio corresponding to quarter two of the
time window (see attachment 1). The lag ratio for this quarter is clearly higher than all the other lag
ratios in the time window. Further, the lag ratios for quarters nine and ten (though not considered
‘final’) strongly suggest that ratios might continue at the level of quarters three to eight.
Revised estimate = 591.
National Centre for Vocational Education Research
15
Attachment 1: Revised estimates for Collection 83
The following graphs depict the pattern of the lag ratios for the estimates that were revised or
considered for revision. The graph shows the lag ratios for the eight quarters in the time window used
in the endorsed model (labelled 1 to 8) and also the two quarters following (labelled 9 and 10).
Horizontal lines are also displayed on the graphs. One represents the average lag as calculated from
the lags in the time window (red line). Where there is another, it represents the average lag as
calculated from the alternative time period used for the revised estimate (green line).
South Australia commencements - initial estimate
1.6
1.4
Lag Ratios
1.2
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
6
7
8
9
10
9
10
Quarters
Lag ratios
Average ratio
South Australia completions - initial estimate
3.0
2.5
Lag Ratios
2.0
1.5
1.0
0.5
0.0
1
2
3
4
5
6
7
8
Quarters
Lag ratios
16
Average ratio
Alternative average
Adjustment notes – March quarter 2015
Northern Territory cancellations/withdrawals - initial estimate
2.0
1.8
1.6
Lag Ratios
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
6
7
8
9
10
Quarters
Lag ratios
Average ratio
Alternative average
Australian Capital Territory commencements - initial estimate
1.4
1.2
Lag Ratios
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
6
7
8
9
10
Quarters
Lag ratios
Average ratio
National Centre for Vocational Education Research
Alternative average
17
Attachment 2: Expired contracts for Collection 83
Although subject to high relative errors, estimates of expired contracts have not been altered because
they are such a small contributor to the in-training estimate. As can be seen from the following
graphs, which depict the pattern of the lag ratios for the estimates of expired contracts, an
alternative way of estimating expired contracts is often unclear.
The graphs show the lag ratios for the eight quarters in the time window used in the endorsed model
(labelled 1 to 8). A horizontal line is also displayed, representing the average lag as calculated from
the lags in the time window (red line).
New South Wales
0.35
0.30
Lag Ratios
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Victoria
0.40
0.35
Lag Ratios
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
Quarters
Lag ratios
18
Average ratio
Adjustment notes – March quarter 2015
Queensland
0.45
0.40
0.35
Lag Ratios
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
South Australia
0.50
0.45
0.40
Lag Ratios
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
19
Western Australia
0.30
0.25
Lag Ratios
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Tasmania
0.40
0.35
Lag Ratios
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
Quarters
Lag ratios
20
Average ratio
Adjustment notes – March quarter 2015
Australian Capital Territory
1.40
1.20
Lag Ratios
1.00
0.80
0.60
0.40
0.20
0.00
1
2
3
4
5
6
7
8
Quarters
Lag Ratios
National Centre for Vocational Education Research
Average Ratio
21
Adjustment note for Collection 82
NCVER examines the quarterly apprentice and trainee estimates produced by the endorsed model in
order to check that the estimates are reasonable. In particular, a decision rule was introduced in
Collection 45 that mandated reviewing all estimates with relative prediction errors of 10% or more.
The goal of the review is to correct for any large bias in estimation that might be caused by changes
in the pattern of reporting practices over time. Note that whilst an estimate might be adjusted for
bias, its associated prediction error is not altered.
The cancellation/withdrawal estimate for the Northern Territory for the September quarter 2014 was
associated with a relative prediction error of 21.5%.
The commencement estimate for South Australia for the September quarter 2014 was associated with
a relative prediction error of 14.3%.
The completion estimate for South Australia for the September quarter 2014 was associated with a
relative prediction error of 25.8%.
The relative prediction errors for expired contracts remained relatively high. Tasmania (33.5%), the
Australian Capital Territory (24.4%), Western Australia (16.4%), and Queensland (14.1%) had the
highest prediction errors while the other states and territories were below 10%.
The contribution of expired contracts to the in-training estimate is usually small both in level and
variation. High relative errors appear to be explained to some degree by the fact that the estimates
are small numbers and therefore any variation is relatively large. Adjustments to the estimates of
expired contracts have little effect on the corresponding estimates of in-training. Consequently, no
alterations to estimates of expired contracts have been made.
Northern Territory
Cancellations/withdrawals for the September quarter 2014
From endorsed model – Estimate = 443; Relative error = 21.5%.
Time window for calculating the average lag factor is from December quarter 2010 to September
quarter 2012.
The revised estimate was calculated by excluding the lag ratios corresponding to quarters two and
three of the time window (see attachment 1). The lag ratios for these quarters are clearly higher than
all the other lag ratios in the time window. Although the lag ratios for quarters nine and ten are not
yet considered ‘final’, they suggest that lag ratios will remain lower than the excluded quarters’
ratios.
Revised estimate = 403.
22
Adjustment notes – March quarter 2015
South Australia
Commencements for the September quarter 2014
From endorsed model – Estimate = 3075; Relative error = 14.3%.
Time window for calculating the average lag factor is from December quarter 2011 to September
quarter 2013.
The lag ratios corresponding to quarters seven and eight are at a higher level than the ratios for the
previous six quarters. However, the lag ratios for quarters nine and ten, whilst not yet considered
‘final’, do not confirm that the high level shown in quarter eight will persist. There is also doubt as to
what level the lag ratios will fall to (if they do fall) as the time window advances. The data presents
no clear basis for adjusting the estimate.
No revision.
Completions for the September quarter 2014
From endorsed model – Estimate = 2628; Relative error = 25.9%.
Time window for calculating the average lag factor is from December quarter 2011 to September
quarter 2013.
The revised estimate was calculated by excluding the lag ratio corresponding to quarter seven of the
time window (see attachment 1). The lag ratio for this quarter is clearly higher than all the other lag
ratios in the time window. Although the lag ratios for quarters nine and ten are not yet considered
‘final’, they strongly suggest that lag ratios will remain lower than the excluded quarter’s ratio and
that an eventual return to the level of quarters one to seven might be possible.
Revised estimate = 2421.
National Centre for Vocational Education Research
23
Attachment 1: Revised estimates for Collection 82
The following graphs depict the pattern of the lag ratios for the estimates that were revised or
considered for revision. The graph shows the lag ratios for the eight quarters in the time window used
in the endorsed model (labelled 1 to 8) and also the two quarters following (labelled 9 and 10).
Horizontal lines are also displayed on the graphs. One represents the average lag as calculated from
the lags in the time window (red line). Where there is another, it represents the average lag as
calculated from the alternative time period used for the revised estimate (green line).
Northern Territory cancellations/withdrawals - initial estimate
2
1.8
1.6
Lag ratios
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
5
6
7
8
9
10
8
9
10
Quarters
Lag ratios
Average ratio
Alternative average
South Australia commencements - initial estimate
1.6
1.4
Lag ratios
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
5
6
7
Quarters
Lag ratios
24
Average ratio
Adjustment notes – March quarter 2015
South Australia completions - initial estimate
2.5
Axis Title
2
1.5
1
0.5
0
1
2
3
4
5
6
7
8
9
10
Quarters
Lag ratios
Average ratio
National Centre for Vocational Education Research
Alternative average
25
Attachment 2: Expired contracts for Collection 82
Although subject to high relative errors, estimates of expired contracts have not been altered because
they are such a small contributor to the in-training estimate. As can be seen from the following
graphs, which depict the pattern of the lag ratios for the estimates of expired contracts, an
alternative way of estimating expired contracts is often unclear.
The graphs show the lag ratios for the eight quarters in the time window used in the endorsed model
(labelled 1 to 8). A horizontal line is also displayed, representing the average lag as calculated from
the lags in the time window (red line).
New South Wales
0.35
0.3
Lag ratios
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Victoria
0.45
0.4
Lag ratios
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
Quarters
Lag ratios
26
Average ratio
Adjustment notes – March quarter 2015
Queensland
0.4
0.35
Lag ratios
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
6
7
8
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
South Australia
0.5
0.45
0.4
Lag ratios
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
Quarters
Lag ratios
Average ratio
Western Australia
0.3
0.25
Lag ratios
0.2
0.15
0.1
0.05
0
1
2
3
4
5
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
27
Tasmania
0.4
0.35
Lag ratios
0.3
0.25
0.2
0.15
0.1
0.05
0
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Australian Capital Territory
1.4
1.2
Lag ratios
1
0.8
0.6
0.4
0.2
0
1
2
3
4
5
Quarters
Lag ratios
28
Average ratio
Adjustment notes – March quarter 2015
Adjustment note for Collection 81
NCVER examines the quarterly apprentice and trainee estimates produced by the endorsed model in
order to check that the estimates are reasonable. In particular, a decision rule was introduced in
Collection 45 that mandated reviewing all estimates with relative prediction errors of 10% or more.
The goal of the review is to correct for any large bias in estimation that might be caused by changes
in the pattern of reporting practices over time. Note that whilst an estimate might be adjusted for
bias, its associated prediction error is not altered.
The cancellation/withdrawal estimate for the Northern Territory for the June quarter 2014 was
associated with a relative prediction error of 19.9%.
The commencement estimate for South Australia for the June quarter 2014 was associated with a
relative prediction error of 12.2%.
The completion estimate for South Australia for the June quarter 2014 was associated with a relative
prediction error of 25.9%.
The relative prediction errors for expired contracts remained relatively high. Tasmania (33.6%), the
Australian Capital Territory (24.6%), Western Australia (17.8%), and Queensland (12.6%) had the
highest prediction errors while the other states and territories were below 10%.
The contribution of expired contracts to the in-training estimate is usually small both in level and
variation. High relative errors appear to be explained to some degree by the fact that the estimates
are small numbers and therefore any variation is relatively large. Adjustments to the estimates of
expired contracts have little effect on the corresponding estimates of in-training. Consequently, no
alterations to estimates of expired contracts have been made.
Northern Territory
Cancellations/withdrawals for the June quarter 2014
From endorsed model – Estimate = 484; Relative error = 19.9%.
Time window for calculating the average lag factor is from September quarter 2010 to June quarter
2012.
The revised estimate was calculated by excluding the lag ratios corresponding to quarters three and
four of the time window (see attachment 1). The lag ratios for these quarters are clearly higher than
all the other lag ratios in the time window. Although the lag ratios for quarters nine and ten are not
yet considered ‘final’, they suggest that lag ratios will remain lower than the excluded quarters’
ratios.
Revised estimate = 443.
National Centre for Vocational Education Research
29
South Australia
Commencements for the June quarter 2014
From endorsed model – Estimate = 3065; Relative error = 12.2%.
Time window for calculating the average lag factor is from September quarter 2011 to June quarter
2013.
The lag ratios corresponding to quarters seven and eight are at a higher level than the ratios for the
previous six quarters. However, the lag ratios for quarters nine and ten, whilst not yet considered
‘final’, do not confirm that the high level shown in quarter eight will persist. There is also doubt as to
what level the lag ratios will fall to (if they do fall) as the time window advances. The data presents
no clear basis for adjusting the estimate.
No revision.
Completions for the June quarter 2014
From endorsed model – Estimate = 4162; Relative error = 25.9%.
Time window for calculating the average lag factor is from September quarter 2011 to June quarter
2013.
The revised estimate was calculated by excluding the lag ratio corresponding to quarter eight of the
time window (see attachment 1). The lag ratio for this quarter is clearly higher than all the other lag
ratios in the time window. Although the lag ratios for quarters nine and ten are not yet considered
‘final’, they strongly suggest that lag ratios will remain lower than the excluded quarter’s ratio and
that a return to the level of quarters one to seven might be possible.
Revised estimate = 3811.
30
Adjustment notes – March quarter 2015
Attachment 1: Revised estimates for Collection 81
The following graphs depict the pattern of the lag ratios for the estimates that were revised or
considered for revision. The graph shows the lag ratios for the eight quarters in the time window used
in the endorsed model (labelled 1 to 8) and also the two quarters following (labelled 9 and 10).
Horizontal lines are also displayed on the graphs. One represents the average lag as calculated from
the lags in the time window (red line). Where there is another, it represents the average lag as
calculated from the alternative time period used for the revised estimate (green line).
Northern Territory cancellations/withdrawals - initial estimate
2.0
1.8
1.6
Lag ratios
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
1
2
7
6
5
4
3
8
9
10
8
9
10
Quarters
Alternative average
Average ratio
Lag ratios
South Australia commencements - initial estimate
1.8
1.6
1.4
Lag ratios
1.2
1.0
0.8
0.6
0.4
0.2
0.0
1
2
3
4
5
6
7
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
31
SA completions initial estimate
2.5
2.0
Lag ratios
1.5
1.0
0.5
0.0
1
2
3
4
5
6
7
8
9
10
Quarters
Lag ratios
32
Average ratio
Alternative average
Adjustment notes – March quarter 2015
Attachment 2: Expired contracts for Collection 81
Although subject to high relative errors, estimates of expired contracts have not been altered because
they are such a small contributor to the in-training estimate. As can be seen from the following
graphs, which depict the pattern of the lag ratios for the estimates of expired contracts, an
alternative way of estimating expired contracts is often unclear.
The graphs show the lag ratios for the eight quarters in the time window used in the endorsed model
(labelled 1 to 8). A horizontal line is also displayed, representing the average lag as calculated from
the lags in the time window (red line).
New South Wales
0.35
0.30
Lag ratios
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Victoria
0.45
0.40
0.35
Lag ratios
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
33
Queensland
0.40
0.35
0.30
Lag ratios
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
South Australia
0.45
0.44
0.43
0.42
Lag ratios
0.41
0.40
0.39
0.38
0.37
0.36
0.35
0.34
1
2
3
4
5
Quarters
Lag ratios
34
Average ratio
Adjustment notes – March quarter 2015
Western Australia
0.30
0.25
Lag ratios
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
6
7
8
6
7
8
Quarters
Lag ratios
Average ratio
Tasmania
0.40
0.35
0.30
Lag ratios
0.25
0.20
0.15
0.10
0.05
0.00
1
2
3
4
5
Quarters
Lag ratios
National Centre for Vocational Education Research
Average ratio
35
Australian Capital Territory
1.20
1.00
Lag ratios
0.80
0.60
0.40
0.20
0.00
1
2
3
4
5
6
7
8
Quarters
Lag ratios
36
Average ratio
Adjustment notes – March quarter 2015