• ## Methodology

To estimate the wine equalisation tax (WET) gap, we use both logistic and linear regressions in a bottom-up, multi-stage regression model.

We estimate the net gap which is the difference between the payable and refundable amounts. For a whole of program view, we include all entities in the WET system.

### Step 1: Calculate a risk score

We consult with the WET compliance area to identify logic and variables for risk selection in ATO compliance activity.

We then calculate a risk score for all WET entities in the population. The score is a key explanatory variable used in subsequent regressions.

### Step 2: Apply logistic regression to estimate the probability of an entity being selected

Apart from the risk score calculated in Step 1, we identify other characteristics of WET entities to help us predict the probability of them being selected for compliance activity.

We then estimate the probability of selection using a logistic regression. This step determines the sample weights for correcting selection bias in the operational audit data.

### Step 3: Apply a weighted logistic regression to estimate the probability of non-compliance of each entity

We analyse the business activity statement (BAS) data of WET entities that were subject to amendment activities, as well as those which were not.

We apply a logistic regression to estimate the probability of a WET entity being found non-compliant as a result of a compliance activity. To adjust for selection bias in the operational audit data, we apply the sample weights calculated in Step 2 to the observations.

### Step 4: Apply a weighted linear regression to estimate the average amended tax as a percentage of WET payable

We analyse the BAS data of known non-compliant WET entities to identify characteristics that would contribute to the prediction of amended tax size. We apply a linear regression to estimate the amended tax size for each entity found to be non-compliant as a share of their WET payable.

The regression is again weighted to adjust for selection bias. The linear regression is then applied to each entity in the population to estimate the potential size of amended tax.

The key difference between Steps 3 and 4 is that Step 3 calculates the likelihood of an entity having a tax gap, while Step 4 calculates the size of each entity's potential amended tax.

### Step 5: Apply the combined results from the 2 weighted regressions to each WET entity

We combine the regression results from Steps 3 and 4 to estimate unreported tax. We estimate total unreported tax by taking the average of the results from 20,000 simulations. This amount includes amendment results.

### Step 6: Apply a non-detection uplift factor and non-pursuable debt

We uplift the estimates before this step to account for non-compliance that is not detected. This ensures the final estimate is not understated.

We also add in the value of non-pursuable debt. This is debt the Commissioner of Taxation has assessed as:

• not legally recoverable
• uneconomical to pursue, or
• unable to be pursued due to another Act.

### Step 7: Consolidate the tax gap estimates

We calculate the gross gap by adding the unreported amounts from Step 5 to the non-detection uplift and non-pursuable debt from Step 6.

We calculate the net gap by subtracting the total amendment amount from the gross gap. Then we add the net gap to the tax paid to estimate the total theoretical liability.

### Summary of the estimation process

Table 2 shows a summary of each step of the estimation process and the results for each year.

Table 2: Applying the methodology – WET gap

Step

Description

2014–15

2015–16

2016–17

2017–18

2018–19

2019–20

1 to 5

Estimate unreported tax (\$m)

7.9

10.0

7.2

14.6

13.9

13.6

6.1

Apply estimate for non-detection (\$m)

5.1

4.9

5.0

5.3

5.3

5.3

6.2

12.7

12.7

12.7

12.7

12.7

12.7

7.1

Gross gap (\$m)

34.6

34.0

34.3

35.5

35.9

35.9

7.2

Amendments (\$m)

9.0

6.4

9.4

2.9

3.9

4.2

7.3

Net gap (\$m)

25.6

27.6

24.9

32.5

32.0

31.6

7.4

Tax paid (\$m)

822

867

858

912

1,014

1,021

7.5

Total theoretical liability (\$m)

848

895

882

944

1,046

1,053

7.6

Gross gap (%)

4.1%

3.8%

3.9%

3.8%

3.4%

3.4%

7.7

Net gap (%)

3.0%

3.1%

2.8%

3.4%

3.1%

3.0%

Find out more about our overall research methodology, data sources and analysis for creating our tax gap estimates.

### Limitations

These limitations are associated with the WET gap estimation:

• There is considerable delay after a financial year ends and the completion of our compliance activities relating to that year. This means gap estimates may subject to revisions for several years.
• The extent of non-detection is not directly observable and extremely challenging to measure. We use a figure based on expert opinion and operational data.
• The calculation of the shadow economy's impact on WET revenue is difficult to measure. For this estimate, we allocate a small amount within the non-detection estimate. This is in keeping with expert opinion that shadow economy activity is infrequent and irregular with a negligible impact on the WET gap.
• The lower coverage levels of compliance activity have resulted in us adopting a pooled regression approach. We draw all data across 2014–15 to 2019–20 together to estimate the coefficients for all years. This implicitly assumes the relationships between variables do not change much over that period. While deriving standalone regression results for each income year would be ideal, it is not feasible due to limited data across the WET population.

### Updates and revisions to previous estimates

Each year we refresh our estimates in line with the annual report. Changes from previously published estimates occur for many reasons, including:

• different methodologies used
• improvements in methodology
• revisions to data

Figure 2 shows the gross and net gaps from our current model compared to our previous estimates.

Figure 2: Current and previous WET tax gap estimates, 2010–11 to 2019–20 The data used in Figure 2 is presented in Table 3 below.

Table 3: Current and previous net wine equalisation tax gap estimates (percentage), 2010–11 to 2019–20

2010–11

2011–12

2012–13

2013–14

2014–15

2015–16

2016–17

2017–18

2018–19

2019–20

2022 program

n/a

n/a

n/a

n/a

3.0

3.1

2.8

3.4

3.1

3.0

2021 program

n/a

n/a

n/a

5.3

3.0

2.8

2.7

3.2

2.9

n/a

2020 program

n/a

n/a

12.8

5.3

3.0

2.8

2.6

3.1

n/a

n/a

2019 program

n/a

6.1

12.6

5.3

3.0

2.6

2.5

n/a

n/a

n/a

2018 program

4.1

4.9

0.5

5.9

0.1

0.5

n/a

n/a

n/a

n/a

2017 program

4.1

4.9

0.5

5.9

0.6

n/a

n/a

n/a

n/a

n/a

2016 program

4.1

4.9

0.5

5.9

n/a

n/a

n/a

n/a

n/a

n/a

2015 program

2.3

4.3

3.3

n/a

n/a

n/a

n/a

n/a

n/a

n/a

Notes: Gap estimates published in 2017–18 and earlier are WET payable gaps only for 2010–11 to 2013–14.