9220.0 - Freight Movements, Australia, Summary, Mar 2001 (Reissue)  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 06/09/2002  Reissue
   Page tools: Print Print Page Print all pages in this productPrint All

TECHNICAL NOTE 2


ADJUSTMENTS APPLIED TO COMPENSATE FOR UNDER-REPORTING


INTRODUCTION

1 Respondents to the FMS 2000 were asked to report on all trips made during a selected two week period (see para. 6 of the Explanatory Notes for further details). Comparisons of data reported for the first week (week 1) of the reporting period with data reported for the second week (week 2) indicated that respondents may have under-reported trips made during week 2 of the reporting period.

2 This technical note outlines the methodological investigations that led to the identification of a statistically significant reporting bias and the methodology used to adjust the survey data.

3 The impact of the adjustments on the estimates is summarised in table T2.2 below.


UNDER-REPORTING

4 To investigate the suspected under-reporting, trip data for each of the 14 days of the reporting period for all 26 fortnights of the survey reference period were aggregated on both a daily basis (i.e. trips made on day 1, trips made on day 2, etc.) and a weekly basis (i.e. trips made during week 1 and trips made during week 2).

5 The daily data clearly showed a strong weekly pattern and the influence of public holiday periods such as Easter and Christmas. There appeared to be a significant difference between week 1 and week 2 data and a statistical test was performed.

6 To properly estimate the week 1/week 2 effect, taking into account public holidays, a model based estimation approach was adopted (using the TRAMO/SEATS package used by Eurostat).

7 Two tests were subsequently performed:

  • one for the effects of the first day/last day of the reporting period; and
  • another for the week 1/week 2 effect (by comparing the averages across the two weeks).

8 The above analysis was run on the daily time series for each of the following variables: laden distance travelled, total distance travelled, weight carried, number of laden trips made, and total number of trips made. Each variable showed strong statistical significance (and consistent direction) as measured by the statistical tests generated by the TRAMO/SEATS package. Therefore, there was justification to adjust the data by the levels of the effects estimated for each of the variables.


IMPUTATION

9 Given the significance of the under-reporting in week 2, a decision was made to compensate for the discrepancy in the estimates. Conceptually, an adjustment for each of the five variables of interest was required (i.e. for laden distance travelled, total distance travelled, weight carried, number of laden trips made and total number of trips made).

10 Imputation of additional trip records for week 2 using donor records from week 1 was the preferred method of adjustment. This method was chosen over other weighting adjustment options because the impact on the data at all levels could be assessed.

11 Initial analysis reviewed the distribution of trips for each of the five variables to identify possible imputation classes. A number of options for imputation classes using one or more of the five variables were evaluated. The best performing classes were for the one variable of laden distance travelled, split into three groups (less than 900 km, 900 km to 2000 km, and 2001 kms and greater).

12 The number of records to be imputed in each of the imputation classes was based on the adjustment levels estimated from the investigations of under-reporting in week 2. The actual adjustment included in the data, via the addition of extra trips, aligned closely with the estimated adjustments required (see table T2.1).

T2.1 PERCENTAGE ADJUSTMENTS TO MAIN DATA ITEMS
Estimated adjustment
Adjustment
levels required
levels realised
Variables
%
%
Laden distance travelled
5.05
5.32
Total distance travelled
4.00
3.88
Weight carried
5.10
5.09
Number of laden trips made
2.30
5.11
Total number of trips made
2.60
2.90


IMPACT

13 The number of imputed records for each imputation class and the impact on the estimates is summarised below (see table T2.2).

T2.2 IMPACT ON ESTIMATES
Tonnes
Tonne-kilometres
Percentage
Value
of total
obtained
Contribution
Value
Contribution
Total
Records
records
from all
from imputed
obtained from
from imputed
records
imputed
imputed
records
records
all records
records
Imputation class (laden
kilometres)
no.
no.
%
'000
'000
'000
'000
Less than 900
181,761
9,425
5.2
592,817
28,959
58,042,271
2,931,621
900 to 2,000
5,603
108
1.9
18,250
334
21,997,521
404,937
2,001 and greater
642
114
17.8
2,854
425
8,334,538
1,286,956
Total
188,006
9,647
5.1
613,921
29,718
88,374,330
4,623,514