ATO logo

Methodology for estimating the FBT gap

Learn the method we use to estimate the fringe benefits tax (FBT) gap.

Last updated 28 October 2026

Method to estimate the FBT gap

To estimate the fringe benefits tax (FBT) gap, we combine judgment-based analysis with logistic and linear regressions in a bottom-up, multi-stage regression model. We selected this method as it draws on our ongoing engagement activities and operational intelligence.

Step 1: Identify employer population and their engagement activities

We identify the employer population through PAYG withholding records. We put these into subpopulations that align to our engagement activities. Key to this is the identification of FBT-registered employers and those not registered for FBT.

Employers not registered for FBT include:

  • employers that declare employee contributions or had reportable fringe benefits amounts in their employees’ PAYG withholding records, but do not lodge an FBT return
  • employers that do not lodge an FBT return, report employee contributions or reportable fringe benefits, but claim motor vehicle expenses in their income tax return
  • all other employers that do not lodge an FBT return but do not fit into the previous categories.

Step 2: Estimate the probability of selection for engagement

For relevant employers identified in Step 1, we identify characteristics that contributed to their probability of being selected for FBT compliance. We use these characteristics to estimate the probability of selection for the broader population via a logistic regression. This step determines the sample weights for correcting selection bias in our FBT compliance data.

Step 3: Estimate the probability of non-compliance

For relevant employers subjected to compliance engagement, we identify characteristics that contributed to their probability of being non-compliant with FBT. We use these characteristics to estimate the probability of being non-compliant with FBT for the broader population via another logistic regression. To adjust for selection bias in our FBT compliance data, we apply sample weights calculated from Step 2 to the observations in Step 3.

Step 4: Estimate unreported amounts

For relevant employers that were non-compliant, we identify characteristics that help predict their amount of unreported FBT. We then apply a linear regression to estimate the potential unreported FBT for the broader population. The data is again weighted to adjust for selection bias using the results of Step 2.

For large employers, we skip Steps 2 and 3, instead accounting for selection bias with a judgment-based correction to average unreported FBT from large employers in our FBT compliance data.

Step 5: Apply a non-detection uplift factor and long-term non-payment

We uplift the estimate from Step 4 to account for non-compliance that is not detected. This ensures the final estimate is not understated.

We also add in the value of long-term non-payment. This is the estimated amount of outstanding FBT liabilities that are unlikely to be ever repaid.

Step 6: Consolidate tax gap estimates and adjust for deductibility

We calculate the gross gap by adding the unreported amounts from Step 4 to the non-detection uplift and long-term non-payment from Step 5. We calculate the net gap by subtracting the total amendment amount from the gross gap. Then we add the net gap to expected collections to estimate the total theoretical liability.

Employers can claim an income tax deduction for the fringe benefits they provide. We determine the size of the forgone income tax deduction that would have been realised if FBT were correctly reported. We adjust the estimates by the relevant income tax rate and subtract from the sum of unreported tax, non-detection and long-term non-payment amounts to produce the net revenue effect amounts.

Summary of estimation process

The steps for the estimation process and the results for each year as a dollar amount and percentage are shown in Table 3.

Table 3: Applying the methodology, FBT gap, 2015–16 to 2023–24

Step

Description

2015–16

2016–17

2017–18

2018–19

2019–20

2020–21

2021–22

2022–23

2023–24

1.1

Population

825,718

848,955

852,996

856,094

898,755

934,717

953,441

964,518

956,733

1.2

Amendments ($m)

24

32

31

26

21

39

22

28

28

2.1

Add estimated unreported amounts ($m)

1,665

1,714

1,707

1,654

1,772

1,881

1,908

1,935

1,884

2.2

Add long-term non-payment ($m)

4

4

5

5

8

8

11

15

16

2.3

Add non-detection ($m)

444

459

456

438

469

506

506

513

500

3.1

Equals gross gap ($m)

2,137

2,208

2,199

2,123

2,270

2,434

2,447

2,491

2,427

3.2

Subtract estimated impact of income tax deductions ($m)

604

602

599

577

621

630

635

619

604

3.3

Equals net revenue effect gross gap ($m)

1,533

1,606

1,600

1,546

1,650

1,804

1,812

1,871

1,823

3.4

Subtract amendments ($m)

24

32

31

26

21

39

22

28

28

3.5

Equals net revenue effect net gap ($m)

1,509

1,575

1,569

1,520

1,628

1,765

1,790

1,844

1,795

3.6

Add expected collections ($m)

4,334

4,163

3,861

3,914

3,919

3,275

3,553

4,161

4,572

3.7

Equals theoretical tax liability ($m)

5,843

5,737

5,430

5,435

5,547

5,039

5,343

6,004

6,368

3.5

Net revenue gross gap (%)

26.2

28

29.5

28.4

29.7

35.8

33.9

31.2

28.6

3.6

Net revenue net gap (%)

25.8

27.4

28.9

28

29.4

35

33.5

30.7

28.2

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

Limitations of FBT gap

The following caveats and limitations apply when interpreting this tax gap release:

  • There is no independent data source which can provide a credible or reliable macroeconomic-based estimate (unlike transaction-based taxes).
  • The data available does not always indicate whether amendments processed are due to our action or taxpayers correcting their own errors.
  • There is a high level of uncertainty around the level of non-detection – the current factor used to account for non-detection is based on factors used in other gaps, and other jurisdictions.

Accounting for the shadow economy

For tax gap purposes we focus on the shadow economy definition which covers activities that are productive and legal but are deliberately concealed to avoid paying taxes or complying with regulations (or both).

In the context of the FBT gap estimate, we attribute the unreported amounts and non-detection associated with employers who do not participate in the FBT system as part of the shadow economy, in Table 4 below. For 2023–24, the shadow economy component accounts for 79% of the overall gross FBT gap.

Table 4: Summary of the impact of the shadow economy as a share of net revenue effect gross gap

Element

2015–16

2016–17

2017–18

2018–19

2019–20

2020–21

2021–22

2022–23

2023–24

Unreported amounts ($m)

1,378

1,423

1,425

1,426

1,487

1,531

1,565

1,591

1,524

Add non- detection ($m)

358

370

371

371

387

398

407

414

396

Equals net gap ($m)

1,736

1,793

1,796

1,796

1,874

1,929

1,972

2,005

1,921

Subtract estimated impact of income tax deductions ($m)

495

493

494

494

515

502

513

501

480

Total shadow economy impact ($m)

1,241

1,300

1,302

1,302

1,358

1,428

1,459

1,504

1,440

Total net revenue gross gap ($m)

1,533

1,606

1,600

1,546

1,650

1,804

1,812

1,871

1,823

Shadow economy as a share of net revenue gross gap (%)

81%

81%

81%

84%

82%

79%

81%

80%

79%

Updates and revisions to previous estimates

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

  • improvements in methodology
  • revisions to data
  • additional information becoming available.

The methodology for 2025–26 is unchanged from last year. The impact of revisions for each published estimate is shown in Figure 3.

Figure 3: Current and previous net revenue effect net gap estimates, 2014–15 to 2023–24

Our previous and current net gap estimates as outlined in Table 5.

This data is presented in Table 5 as a percentage.

Table 5: Current and previous FBT gap estimates (%), 2015–16 to 2023–24

Program year

2015–16

2016–17

2017–18

2018–19

2019–20

2020–21

2021–22

2022–23

2023–24

2026

25.8

27.4

28.9

28

29.4

35

33.5

30.7

28.2

2025

25.3

27.1

28.6

27.8

29.4

35.2

33.2

30.4

n/a

2024

26.5

28.5

30.2

29.6

31.2

36.3

34.8

n/a

n/a

2023

28.3

23.4

26.1

24.1

20.1

28.2

n/a

n/a

n/a

2022

28.4

21.1

22.1

22

20.3

n/a

n/a

n/a

n/a

2021

26.7

22

22.4

22.6

n/a

n/a

n/a

n/a

n/a

2020

26.8

22

21.2

n/a

n/a

n/a

n/a

n/a

n/a

QC105715