Statistical Clearing House

Agricultural Finance Survey, Financial Year 1999/2000

---

• Overview
• Quality Framework
• A. General Information
• B. User Requirements, Survey Objectives and Budget
• C. Alternative Sources of Data
• D. Data Outputs
• E. Questionnaire Design and Data Collection
• F. Respondent Load
• G. Population / Frame
• H. Sample Design *** You are here ***
• H1Sample design
  
• H2 Sample stratification
  
• H3 Sample allocation
  
• H4 Sample size
  
• H5 Expected live and in scope units
  
• H6 Response rate
  
• H7 Representativeness of respondents
  
• H8 Allowances for non-response
  
• H9 Respondent burden across surveys
• I. Estimation and Analysis

H. Sample Design

---

H1 Sample design

What sample design will be used, eg, census, stratified simple random sample? (H1)
Note. Should include the sample selection strategy, and, if the sample design is multi-stage or multi-phase (or both), the sample selection strategy used in each stage/phase.

Stratified synchronised random sample

The CEd sector is defined by EVAOs that are over a certain limit dependent on State and industry (ie. atypically large). LBUs are automatically CEd. The CEd sector comprises approximately 10% of the total sample.

The rate of rotation is planned at one third of the sample. The achieved rate, however, is nearly 60%, mainly due to stratum flipping (ie units selected but have found to have moved strata), ie. 33% rotate as planned - the remaining rotation is unplanned.

Important multi-state MUs are separately stratified. Importance is measured by EVAO contribution. We attempt to get about 75 -80% of total EVAO in the multi-State strata. Previous empirical work has shown that the remaining ones can be treated as single State MU without generating any significant bias. Note that some of these can then be CEd at the Industry level.

If a census, go to H4

Return to top

H2 Sample stratification

What stratification has been used in the sample design? (H2)
Note. Should include: stratification variables; stratification boundaries (noting any strata which are completely enumerated); reasons stratification strategy adopted; and, if multi-stage/phase sampling is being used, report for each stage/phase.

Stratification used:
State (ACT combined with NSW),
Industry (13 industry classes - state dependent - based on ANZSIC class),
Size (estimated value of agricultural operations).

Return to top

H3 Sample allocation

How has the sample been allocated to strata? (H3)
Note. Should include: allocation formulae (if used), and from where data referenced by formulae obtained; minimum and maximum stratum sample size constraints; reasons why the allocation strategy was adopted; and, if multi-stage/phase sampling is being used, report for each stage/phase.

Sample is optimally allocated using turnover as a design variable.
The allocation aims to meet RSE requirements at National (2%), State (5% - 7%) and Industry (10% - 15%) levels.

Minimum sample sizes:
Size 1 strata - 8 (note stratum boundaries are based on EVAO levels and these vary across the industries)
Other strata - 6

Allocations are adjusted to account for expected non-response

A minimum response of 2 per stratum is required for estimates.

Return to top

H4 Sample size

What is the expected total sample size to be approached and how has it been set? (H4)
Note. Should include: any constraints which have been imposed (eg cost); if multi-stage/phase sampling is being used, should report for each stage/phase; if the survey is a census, then the total sample size should be the same as the total number of units on the frame reported in G4; and, if quota sampling is used, should indicate the number of businesses that agency will attempt to contact.

Return to top

H5 Expected live and in scope units

Of the total sample size reported in H4, how many units are expected to be live and in scope? (H5)
Note. Should include how these estimates were determined and, if the sample is multi-phase, should include these quantities for each phase.

Return to top

H6 Response rate

Of the total sample size reported in H5, how many units are expected to respond? (H6)
Note. Should include: how these estimates were determined; target relative standard error (RSE) levels on selected data output items (if any) and reasons for choice; if the sample is multi-phase, should indicate these quantities for each phase; and, if quota sampling is used, should indicate the size of the desired sample (this should be a subset of H5).

Return to top

H7 Representativeness of respondents

What evidence is there that the expected respondents will be representative of non-respondents? (H7)

AFS achieves a high response rate and the effect of nonresponse should be minimal. To the extent that we are interested in optimising mean square error, we have not found evidence to suggest that using implicit mean imputation in the sampled sector causes undue bias.

Return to top

If a census, go to I1

H8 Allowances for non-response

What allowances have been made for expected non-response and frame deficiencies in the sample design? (H8)
Note. Should include allowances made for: refusals, non-contacts, expected defunct units (ie units no longer in business), and expected out of scope units.

Allowance is made for a non-response (for various reasons) of approximately 15% by increasing the sample size.

Minimum sample sizes are set.

Return to top

H9 Respondent burden across surveys

What consideration has been given to ensuring respondent burden is spread evenly by controlling sample selection across surveys, or (for repeating surveys) over survey cycles? (H9)

Considerations include:
The use of synchronised sampling
For the non-CEd sector, the target is a maximum of 3 years in the survey. Units selected for more than 3 years are reviewed during sample design and selection and there may be some manual adjustment of rotation intervals to alleviate respondent burden problems.

Return to top