|Page tools: Print Page Print All|
23.12 Units in the ABSMP are maintained via a combination of personal profiling, mail profiling and targeted unit survey activity. The large, more complex units are profiled annually. The balance of the ABSMP is maintained on a rotating basis within a reasonable timeframe.
23.13 For further information on the ABS economic statistics units model, refer to Information Paper: Improvements in ABS Economic Statistics [Arising from The New Tax System] (cat. no. 1372.0).
THE COMMON FRAME
23.14 Each quarter a snapshot of the ABS Business Register, referred to as the common frame, is taken. The snapshot of the register is used as the frame for all sub-annual business surveys covering the reference quarter for which the snapshot was extracted. This common timing ensures that populations and sampling information are consistent across all surveys with a common reference period.
NEW BUSINESS PROVISIONS
23.15 New business provisions are adjustments made to survey estimates to account for births and deaths of businesses that have occurred up to the end of the survey reference period but which are not reflected on a survey frame (i.e.births and deaths which occur after the snapshot is taken but before the end of the survey reference period). The calculation of the new business provisions makes assumptions about the number of new ABN registrants which will result in births of ABN units on the ABS Business Register, and the average contribution of these birthed units to the item being estimated. The methodology assumes that the contribution of a new business is the same as the average contribution to estimates of like units already on the frame.
SCOPE AND COVERAGE OF ABS BUSINESS SURVEYS
23.16 The scope of ABS business surveys varies across collections. Most surveys using the ABS Business Register as a frame have either an industry orientation (e.g. monthly retail surveys) or an economy-wide orientation (e.g. the Economic Activity Survey and the labour-related business surveys). The scope of surveys that use the business register as a frame is restricted by the scope and coverage of the register itself (see discussion above). Surveys with broader or different scope are required to either supplement the business register for their frame, or use frames that have been constructed independently of the register by using relevant alternative data sources.
23.17 While the scope of ABS labour-related business surveys varies across collections, most are restricted to employing businesses (those which have registered for the ATO's PAYGW Scheme). However, the following groups of employing businesses are generally excluded from labour-related business surveys:
STATISTICAL UNITS USED IN ABS BUSINESS SURVEYS
23.18 The statistical units used in ABS business surveys vary across collections.
23.19 For business surveys using the ABS Business Register as a survey frame, the sampling unit most commonly used is the ABN unit for businesses with simple structures, and the TAU for businesses with complex structures. For these surveys, the sampling unit (ie. the ABN unit or the TAU) is often the same as the other statistical units (e.g. reporting and collection units) used in the survey. However, this is not always the case and in some surveys the statistical units differ.
23.20 For business surveys using alternative frames, the sampling unit usually corresponds to one of the economic units defined on the ABS Business Register.
23.21 For labour-related business surveys, a variety of statistical units are used. With the exception of the Industrial Disputes collection, all labour-related business surveys use the ABS Business Register as their frame. An independent frame is constructed for the Industrial Disputes collection. The sampling unit most commonly used for labour-related business surveys is the ABN unit (for businesses with simple structures), and the TAU (for businesses with complex structures) for each state of operation of the ABN unit/TAU.1. The collection and reporting unit used in the labour-related business surveys usually corresponds to the sampling unit. Where organisations are unable to provide data at the TAU/state level, special reporting units are created to collect information from the level within the organisation structure at which it is available. These special reporting units comprise either an aggregration of TAU/state units into larger units or a disaggregration into smaller units. In cases where a TAU/state unit is disaggregrated into a large number of similar small units, a sub-sample of these units is selected in order to reduce the reporting load on the providers.
1. Prior to the September 2002 quarter, most labour-related business surveys used the management unit/state (MUS) as the statistical unit. In most cases, ABN unit/TAU accords with the management unit.
COLLECTION METHODS USED IN ABS BUSINESS SURVEYS
23.22 Most ABS business surveys use either the mail-out/mail-back or the telephone interview collection methodology. A number of surveys use administrative by-product data. Please refer to Chapter 16 for further explanation of different collection methodologies.
23.23 Most business surveys, including the labour-related business surveys, have intensive follow-up procedures for non-response. Follow-up normally involves reminder letters and may also involve telephone contact. 'Priority' intensive follow-up is used for a number of surveys; this involves targeting the following types of non-responding units:
SAMPLE DESIGN USED IN ABS BUSINESS SURVEYS
23.24 Two types of business surveys are undertaken by the ABS: censuses, and probability sample surveys. Most ABS business surveys (whether they be censuses or sample surveys) use the ABS Business Register to create their survey frame. With the exception of the Industrial Disputes collection, all labour-related business surveys construct their frame from the ABS Business Register. An independent frame is constructed each month for the Industrial Disputes collection.
23.25 Once a survey frame has been compiled, whether it be from the ABS Business Register or another source, information is collected from certain businesses on the frame. In a census, information is collected from all of the units on the frame. In the case of probability sample surveys, information is collected from a random sample (or subset) of units on the frame.
23.26 When a sample is to be selected for an ABS business survey, a survey frame must first be drawn from the common frame (see paragraph 23.14). The survey frame is then divided (stratified) into groups (strata) with similar characteristics. The stratification variables used in the labour-related business surveys include state, sector (public/private), industry and employment size. Once the frame has been stratified, a small number of strata are completely enumerated, while for each of the remaining strata an equal probability sample is taken. Surveys using multi
23.27 There are various constraints placed on sample selection. For most labour-related business surveys, sample selection is constrained by ensuring that a portion of the sample that is not completely enumerated is rotated, and that small businesses will be in the sample for no more than 12 successive quarters (for quarterly surveys other than the Labour Price Index). Some surveys are further constrained by ensuring that there is either minimal or maximal overlap with other surveys.
23.28 The level of sample rotation within surveys, and sample overlap between surveys, is controlled through the use of a system of synchronised sampling. The technique relies on assigning a permanent random number between 0 and 1 to each unit on the ABS Business Register. Samples selected from survey frames constructed from the ABS Business Register are restricted to those units with random numbers that lie in a certain interval. Rotation control is achieved by moving the interval to the right, and overlap control is achieved by constraining the selection interval for each survey to move within a specific range within the larger (0,1) interval.
23.29 Sample sizes vary across ABS business surveys. In determining the minimum required sample size for each survey, the following factors are taken into consideration:
23.30 The sample is allocated across the survey strata using techniques to minimise the variance of the estimates being produced for a fixed cost, or to minimise the cost for a fixed variance. This process is referred to as optimal allocation.
ESTIMATION METHODS USED IN ABS BUSINESS SURVEYS
23.31 Once survey data have been obtained, they must be weighted up to values that represent the whole (target) population. As discussed in Chapter 16 the estimation procedure is essentially the application of weights to the individual survey records. For ABS business surveys, the values of these weights are determined with respect to one or more of the following factors:
23.32 For censuses, aggregates need to be adjusted to account for non-response and problems with the frame.
23.33 Two main types of estimates are used in surveys constructing their frame from the ABS Business Register, namely number-raised estimation and ratio estimation. While some surveys rely solely on number-raised estimation, most will use ratio estimation in some strata. The labour-related business surveys use stratum-by-stratum ratio estimation in strata where the benchmark information is known and sampling efficiencies achievable are greater than with number-raised estimation; for strata where benchmark information is not available, number-raised estimation is used. Number-raised estimation and ratio estimation are discussed further in Chapter 16.
23.34 The post-stratification of the sample prior to estimation is common practice for ABS business surveys, to improve the estimation process. Post-stratification is used in the labour-related business surveys. For further explanation of post-stratification estimation see Chapter 16.
23.35 Editing is used in ABS business surveys to correct a number of non-sampling errors such as those introduced by misunderstanding of questions or instructions, miscoding, non-availability of data, incorrect transcription, non-response and non-contact. Editing can be performed both on the values collected from respondents and on the accumulated survey results as part of the estimation process.
23.36 Some labour-related business surveys use 'significance editing' to reduce the editing load for the survey. Significance editing involves assessing each survey value to be edited against how greatly the survey estimate will be affected by using the unedited value. Only those values which will significantly affect the survey estimate are then edited.
23.37 Adjustments for non-response are made in the estimation process for all business surveys. There are three categories of non-response for ABS business surveys: partial non-response; complete non-response; and refusal. The extent to which values are imputed depends upon the amount and the quality of data already provided. Partial imputation is used when the respondent has provided some data. Complete imputation is used when no data have been provided, or the respondent has refused to participate in the collection.
23.38 There are two sources from which imputed values can be derived for business surveys. The first is data provided by each unit or by a subset of units within a stratum. This may be data previously provided, or current data where partial non-respondents are able to provide some information. One specific example of this type of imputation is 'Beta imputation'. Beta imputation is an automatic imputation procedure that involves estimating the missing values by applying a growth rate to the most recently reported data for these units, provided that data have been reported in either of the two previous quarters. Collections run on an infrequent or one-off basis will not be able to use data previously provided for the imputation of a current missing value.
23.39 The second method derives imputed values by generating a mean value for all missing data items. This method uses the Live Respondent Mean, which is used to impute for units which are known or taken to be alive, but for which no other information is available or useful. Records imputed using the 'mean' do not contribute to estimates of variance.
23.40 Another imputation technique which has been used, in the completely enumerated sector of infrequent surveys where no historical information is available, is a version of the hot-deck method where a non-responding unit gains the reported values of a responding unit which has similar characteristics.
23.41 For complete non-response and refusals in completely enumerated strata, all data items for the unit are imputed, preferably using previously provided data. For complete non-response and refusals in other strata, the unit can have data items imputed using the Live Respondent Mean, although imputation based on previously provided data is preferred. Refer to Chapter 16 for further information on non-response adjustment.
23.42 Two main methods for the treatment of outliers are used in ABS Business Surveys, namely Surprise Outliering and Winsorization. For further explanation of these methods see Chapter 16.
NEW BUSINESS PROVISIONS
23.43 New business provisions are used in the estimation process to allow for births and deaths of businesses that have occurred up to the end of the survey reference period but which are not reflected on a survey frame. The provisions are determined following the creation of the common frame, but are not applied until estimation. For further information, refer to paragraph 23.15.
23.44 The response rate for ABS business surveys is calculated as the number of responding units divided by the total number of live units in the sample.