applying behavioural science in tax administration a ......2017/06/13 · % behavioural change...

Applying Behavioural Science in Tax Administration A Summary of Lesson Learned

June 8 2017

Seán Kennedy Senior EconomistRevenue

Applying Behavioural Science in Tax Administration – A Summary of Lessons Learned

Opinions expressed in this presentation are the views of the author and may not reflect the views of the Office of the Revenue Commissioners. The author alone is responsible for the conclusions.

Structure

PART I – BEHAVIOURAL SCIENCE IN REVENUE1. Our Experience2. Current Research3. Which Insights Drive Behaviour?4. Summary Findings

PART II – FOUR EXPERIMENTS1. Deterrence | Pub Licenses2. Personalisation | Post-it3. Simplification & Salience | Non-Filers 20144. Social Norms | Non-Filers 2016

PART III – LESSONS LEARNED


Our Experience

Application of psychological insights (using letters) to economics

Endorsement Internationally and in Ireland

In general, many nudges; little measurement. Our approach has been empirically focused (observation leads, theory follows)

Empirical demonstrations are persuasive. Where our trials have shown a positive behavioural change, they have been scaled nationwide

While RCTs are scientific, they are also highly specific and not always feasible

Our experience is that endorsement and logistics are often harder than experimental design

Understanding what does not work important as understanding what does (file drawer problem). One of our most striking results is that our evidence on social norms conflicts with similar trials from the UK (but not from the US)


Our current research paper summarises lessons from 20 RCTs conducted by Revenue over the past 6 years

Revenue was the first Government Department in Ireland to implement RCTs to test the application of BE. We have now built up a comprehensive picture of taxpayer behaviour in Ireland

Audit (and other interventions) are effective but expensive and time consuming. Could targeted treatments be more efficient in improving compliance (filing, reporting and paying)

Current Research


Which Insights Drive Behaviour?


Detterence Personalisation Simplification Social Norms

(5) (2) (8) (5)

+8.0% +4.0% +3.3% -1.6%

1. Deterrence: Strategies dissuade taxpayers from noncompliant behaviour2. Simplification & Salience: Simple presentation and drawing attention to details3. Personalisation: Personalised correspondence may improve behaviours4. Social Norms: The behaviour of others can influence an individuals choice

(19) 35.0%

(20) 6.1%

(7) -1.1%

(8) 6.9%

(4) 10.5%

(14) -2.0%

(18) 23.1%

(10) 6.2%

(13) 5.0%(9) 2.6%(1) 1.0%

(15) -4.0%(11) -4.3%

(16) 2.0% (5) 0.3%

(17) 17.0%

(12) -1.3%

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

35%

4 6 8 10 12 14

DeterrenceSimplification & SalienceSocial NormPersonalisation

% Behavioural Change (Mean)Insights

Impact

Larger Sample Size


Summary Findings, by Trial

8.0%

(5 Trials)

3.3%

(8 Trials)

-1.6%

(5 Trials)

4.0%

(2 Trials)

-3%

-1%

2%

4%

6%

8%

10%

5 7 9 11 13 15


% Behavioural Change (Mean)

Larger Sample Size

Insights

Impact

(19) 35.0%

(20) 6.1%

(7) -1.1%

(8) 6.9%

(4) 10.5%

(14) -2.0%

(18) 23.1%

(10) 6.2%

(13) 5.0%(9) 2.6%(1) 1.0%

(15) -4.0%(11) -4.3%

(16) 2.0% (5) 0.3%

(17) 17.0%

(12) -1.3%

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

35%

4 6 8 10 12 14


% Behavioural Change (Mean)Insights

Impact

Larger Sample Size


Summary Findings, by Trial

Deterrence should positively influence taxpayers (Slemrod, 2007) but empirical evidence seems mixed. Wenzel (2004) finds that when norms in favour of compliance are strong, deterrence effects are small, but they become more important when norms are weak

All Publican’s required to hold excise licence issued by Revenue. Renewal letters to all publicans issue in September

Treatment group (400): Randomly assigned to receive revised letter with deterrence (‘unlicensed trading is an offense’) and social norm (‘majority license holders renew on time’)

Control group (7,800): Remainder received standard renewal letter

Target: Improve renewal rates (and timing)

Deterrence (Pub Licenses)


On Time Renewal Rates (1st October): Control 29.4%, Treatment 35.5% At end of licensing year, renewal rates are 97% and 89% respectively While 6% may seem small, it derives from modest changes to a letter

89%

20%

30%

40%

50%

60%

70%

80%

90%

100%

01-Oct-11

01-Nov-11

01-Dec-11

01-Jan-12

01-Feb-12

01-Mar-12

01-Apr-12

01-May-12

01-Jun-12

01-Jul-12

01-Aug-12

01-Sep-12

% Renewing Excise License

Date of Renewal

97%

89%

20%

30%

40%

50%

60%

70%

80%

90%

100%

01-Oct-11

01-Nov-11

01-Dec-11

01-Jan-12

01-Feb-12

01-Mar-12

01-Apr-12

01-May-12

01-Jun-12

01-Jul-12

01-Aug-12

01-Sep-12

% Renewing Excise License

Date of Renewal


Deterrence (Pub Licenses)

Experimental Design Illustrative Example

SAMPLE 2,026 survey forms issued to

SMEs

TREATMENT300 (15%) received survey form and covering letter with short

personalised handwritten post-it note requesting completion

CONTROL1,726 (85%) received survey form and covering letter only

Dear Joe Bloggs,

Please take a few minutes to complete this for us.

Thank you.

Seán KennedyTax Official

Literature suggests seemingly insignificant post-it leads to much higher responses Five specific personal features included in each note

Personalisation (Post-it)


0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

55%

60%

65%

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65

Treatment (300) - PersonalisedNote Attached

Control (1726) - No PersonalisedNote Attached

16%

60%

43%

22%

42%

30%

Reminder LettersResponse Rate

Days



Treatment (300): Response rate is 60% over the period

Response Rates, by Day

Econometric Modelling, by SME taxpayer

Simple OLS (1) regression shows receiving post-it increases response rate by 14.7 percentage points on average

When controls are included, estimate is revised to 13.9

Econometric modelling confirms responses received more quickly

Table 1: OLS, Logit and Probit Estimates of the Effect of Post-it Notes on Survey Responses

All Responses Responses within Four Weeks

OLS (1)

OLS (2)

Logit (1) Probit (1) Logit (2) Probit (2) OLS (3)

Received Post-it 0.147*** 0.139*** 0.139*** 0.139*** 0.175*** 0.176*** 0.176***

(-0.0309) (-0.0314) (-0.0312) (-0.0312) (-0.0298) (-0.0298) (-0.0298)

Controls No Yes Yes Yes Yes Yes Yes

Selected Covariate Estimates

Larger Size

0.003 0.003 0.003 -0.011 -0.011 -0.011

(-0.0278) (-0.0276) (-0.0275) (-0.0233) (-0.0232) (-0.0232)

Manufacturing

0.118* 0.118* 0.119* 0.096 0.095 0.095

(-0.0594) (-0.0592) (-0.0593) (-0.0522) (-0.0524) (-0.0524)

Dublin

-0.072* -0.072* -0.073* -0.04 -0.041 -0.041

(-0.032) (-0.0319) (-0.0319) (-0.0272) (-0.027) (-0.027)

Sole Trader

0.028 0.028 0.028 0.005 0.004 0.004

(-0.0243) (-0.0241) (-0.0241) (-0.0201) (-0.0201) (-0.0201)

Sample size

2,013 2,013 2,013 2,005 2,013 2,005

Table 1: OLS, Logit and Probit Estimates of the Effect of Post-it Notes on Survey Responses

All Responses Responses within Four Weeks

OLS (1)

OLS (2)

Logit (1) Probit (1) Logit (2) Probit (2) OLS (3)

Received Post-it 0.147*** 0.139*** 0.139*** 0.139*** 0.175*** 0.176*** 0.176***

(-0.0309) (-0.0314) (-0.0312) (-0.0312) (-0.0298) (-0.0298) (-0.0298)

Controls No Yes Yes Yes Yes Yes Yes

Selected Covariate Estimates

Larger Size

0.003 0.003 0.003 -0.011 -0.011 -0.011

(-0.0278) (-0.0276) (-0.0275) (-0.0233) (-0.0232) (-0.0232)

Manufacturing

0.118* 0.118* 0.119* 0.096 0.095 0.095

(-0.0594) (-0.0592) (-0.0593) (-0.0522) (-0.0524) (-0.0524)

Dublin

-0.072* -0.072* -0.073* -0.04 -0.041 -0.041

(-0.032) (-0.0319) (-0.0319) (-0.0272) (-0.027) (-0.027)

Sole Trader

0.028 0.028 0.028 0.005 0.004 0.004

(-0.0243) (-0.0241) (-0.0241) (-0.0201) (-0.0201) (-0.0201)

Sample size 2,013 2,013 2,013 2,013 2,005 2,013 2,005



Simplification (Non-Filers)

Self-assessed Income Tax compliance rate is ~90% annually in Ireland

In 2014, conducted Non-Filers Project to evaluate impact of tailoring letters on the proportions of non-filing Income Tax payers who file a tax return

Previously, policy was to issue automated reminder letter to all non-filers

1. Control Group (79,000): Issued bulk letter in March 20142. Treatment 1 (914): Random assignment to receive TPI3. Treatment 2 (912): Random assignment to shorter letter & TPI


25%

0%

5%

10%

15%

20%

25%

30%

35%Total 2014 (79,061)

% Filing Income Tax Return

Days Post-Issue

25%

30%

0%

5%

10%

15%

20%

25%

30%

35%Total 2014 (79,061)

Group 1 2014 (TPI; 914)

Group 2 2014 (T + TPI; 912)

% Filing Income Tax Return

Days Post-Issue

Treatment 2 (912): For those assigned shortened letter with TPI, now significant increase in filing of 5 percentage points…


Simplification (Non-Filers)

% Filing (Form 11), 185 days post letter issue

Social Norms (Non-Filers)

Internationally evidence that the behaviour of others strongly influences an individual’s choices. From tax perspective, perception that majority are compliant may make people less willing to evade taxes themselves

Jan 2016, Revenue tested the impact of a localised social norm

Non-filers in 3 counties (Donegal, Meath and Limerick) randomly assigned to receive standard or localised letter (‘9 in 10 filed in your local area’)

In total, 11,246 letters issued, 50% of which were localised


0%

2%

4%

6%

8%

10%

12%

3 7 11 15 19 23 27 31

Donegal - Standard Letter (1,607)

% Filing

Days Post-Issue

0%

2%

4%

6%

8%

10%

12%

3 7 11 15 19 23 27 31

Donegal - Standard Letter (1,607)

Donegal - Local Letter (1,606)

Limerick - Standard Letter (1,864)

Limerick - Local Letter (1,864)

Meath - Standard Letter (2,152)

Meath - Local Letter (2,152)

% Filing

Days Post-Issue


Social Norms (Non-Filers)

Summary: Social norms combined with geographic specificity do not significantly impact on filing compliance for any of the counties tested

Limerick Treatment (1,864): Again, no observable difference with social norm

Lessons Learned

Our research shows wording, design and tone matters for influencing compliance. Even seemingly insignificant changes to correspondence can significantly changes behaviour

Lessons learned primarily from letters but they can, and should, be considered in any communication (including electronic channels)

Conducting RCTs not just about careful experimental design. Challenges of logistics and organisational endorsement often more challenging

While RCTs are scientific, they are also highly specific and not always feasible

Empirical demonstrations are persuasive. Where our trials have shown a positive behavioural change, they have been scaled nationwide

Understanding what does not work important as understanding what does (file drawer problem). One of our most striking results is that we find no evidence that social norms impact behaviour (different from similar trials in UK but similar to some in US)


applying behavioural science in tax administration a ......2017/06/13 · % behavioural change...

Documents