Evaluation of an intervention to
increase online filing of individuals’ tax
returns
Peter Lumb
September 2009
Contents
• Self Assessment
• SA316 pilot
• Summary results
• Logistic regression to predict adoption of online filing
• Summary and Conclusions
SA316 online filing pilot | September 2009 | 2
Self Assessment – key facts
• Income Tax returns for Individuals, Partnerships and Trusts
• 9.6 million returns for 2007/08
• HMRC has had targets for increasing electronic filing for
several years
• For tax year 2007/08, 69% of returns filed by 31 Jan 2009
deadline were by Internet
• Internet filing saves administrative costs and is liked by the
taxpayers who use it
SA316 online filing pilot | September 2009 | 3
Self Assessment: how taxpayers are asked to file
• Customers who previously filed on paper normally sent a
paper return and guidance. Two types of paper return:
• Short tax return (STR) (four sides A4 paper plus guidance)
• Full/ main tax return: for more complex affairs/ income sources:
longer with supplementary schedules for infrequent income types
• Customers who previously filed by online or through an
advisor/ accountant sent a 2 side letter: Notice to Complete
a return (SA316 form)
• SA316 is not a tax return itself. Advises customers how to file online
or via a tax agent, or otherwise contact HMRC for a paper form.
SA316 online filing pilot | September 2009 | 4
SA316 pilot
• Research question: Does sending an SA316 – instead of a
paper return and associated guidance – to unrepresented
taxpayers previously filing on paper encourage online filing?
• Pilot group: 10,000 individuals who filed on paper for
2005/06: send SA316 notice for 2006/07 tax year – instead
of paper return
• Control group: a further 1,000 taxpayers
• Selected from one office for practical reasons
SA316 online filing pilot | September 2009 | 5
Summary results
• Filing method of taxpayers in pilot and control groups for
2006/07 analysed
• Analysed separately for people sent a Short tax return and
those sent a full/main tax return for previous year (2005/06)
SA316 online filing pilot | September 2009 | 6
Short tax return filers (2005/06) – filing for 2006/07
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Full return Short return Internet Very late or Other
no filing
Control group Pilot group
SA316 online filing pilot | September 2009 | 7
Full return filers (2005/06) – filing for 2006/07
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Full return Short return Internet Very late or Other
no filing
Control group Pilot group
SA316 online filing pilot | September 2009 | 8
Logistic regression: to predict switching to online filing
• Aim:
• To gain a better understanding of factors associated with adopting
online filing
• To build a statistical predictive model that can be used to target
taxpayers most likely to adopt online fling
• Process:
• Exploratory analysis
• Model building and results
SA316 online filing pilot | September 2009 | 9
Adoption of online filing: previous year’s return type
30%
20%
10%
0%
Short return Full return Short return Full return
Control Pilot
SA316 online filing pilot | September 2009 | 10
Adoption of online filing: male/female, control/pilot splits
30%
20%
10%
0%
Male Female Male Female
Control Pilot
SA316 online filing pilot | September 2009 | 11
Adoption of online filing: employee and control/pilot
30%
20%
10%
0%
Employee not Employee Employee not Employee
Control Pilot
SA316 online filing pilot | September 2009 | 12
Adoption of online filing: employee (directors only)
30%
20%
10%
0%
Director Not director Director Not director
Control Pilot
SA316 online filing pilot | September 2009 | 13
Adoption of online filing: self employed (sole trade)
30%
20%
10%
0%
Sole trader Not trader Sole trader Not trader
Control Pilot
SA316 online filing pilot | September 2009 | 14
Adoption of online filing: land and property income
30%
20%
10%
0%
Landlord Not Landlord Landlord Not Landlord
Control Pilot
SA316 online filing pilot | September 2009 | 15
Adoption of online filing: free software applicable
30%
20%
10%
0%
No Yes No Yes
Control Pilot
SA316 online filing pilot | September 2009 | 16
A
B -
- H Su
ap cce
0%
10%
20%
30%
40%
C py ssf
- H u
D Sur ou l St
- s a
E On vivin em rt
F - Fl th g S ate
- C ou e s
re ris Bre ingl
G dit- hin ad es
- G hu g L
ilt- ng Fam ine
H edg ry F ilie
- M ed am s
I - id- Lif ilie
M life es s
J - ode Af tyle
Ad st flue s
M
K van id nce
L - A ci -ye
- W g ng a
e r
M ea ing Sta s
- E lth W tu
ld y R ork s
er e
ly tire ers
De m
No priv ent
t c ati
No las on
t m sifie
at d
ch
ed
Adoption of online filing: Financial Strategy Segments
SA316 online filing pilot | September 2009 | 17
Adoption of online filing: Marital Status
40%
30%
20%
10%
0%
Widowed
Widowed
Single
Single
Not available
Not available
Divorced/ Separated
Divorced/ Separated
Married
Control Married Pilot
SA316 online filing pilot | September 2009 | 18
Adoption of online filing: previous year’s filing date
1.0
0.8
30 Sept. filing deadline
0.6
0.4
0.2
0.0
31 May 06 30 Sep 06 31 Jan 07
SA316 online filing pilot | September 2009 | 19
Adoption of online filing: Age and Sex
Control Pilot
1.0
0.5
0.0
20 40 60 80 100 20 40 60 80 100
Male Female
SA316 online filing pilot | September 2009 | 20
Adoption of online filing: Income
Control Pilot
1.0
0.5
0.0
0 10 40 100 200 350 0 10 40 100 200 350
Income (000s) (square root scale)
SA316 online filing pilot | September 2009 | 21
Logistic regression: model
• Good fit achieved
• (Hosmer-Lemeshow statistic p value 0.92)
• Good predictive capability
SA316 online filing pilot | September 2009 | 22
Predicted probabilities by age, control and pilot groups
Full return in 2005/06
0.8
0.6
0.4
0.2
0.0
20 30 40 50 60 70 80 90
Short return in 2005/06
0.8
0.6
0.4
0.2
0.0
20 30 40 50 60 70 80 90
SA316 online filing pilot | September 2009 | 23
Model performance
0.6
0.5
0.4
0.3
Overall proportion switching
0.2
0.1
0
1 2 3 4 5 6 7 8 9 10
Taxpayers ordered by predicted likelihood of switching to online filing (deciles)
SA316 online filing pilot | September 2009 | 24
Logistic regression: key results – on adoption of online
filing – pilot group
• Pilot cases that could use HMRC software – 4.2 times more
likely
• Employees in pilot group – 2.6 times more likely
• Both employee and HMRC software applicability – 10.9
times more likely
• Directors less likely to switch than other employees
SA316 online filing pilot | September 2009 | 25
Logistic regression: key results – on adoption of online
filing – control group
• Pilot cases that could use HMRC software in control group
– about half as likely as other control group cases
• Employees in control group – about half as likely
• Both employee and can use HMRC software – around a
quarter as likely to adopt online filing as other customers in
control group
SA316 online filing pilot | September 2009 | 26
Logistic regression: key results – on adoption of online
filing – all customers (both pilot and control)
• Single and also divorced/separated individuals – less likely
to switch
• Men slightly more likely than women to switch to online
filing, depending on return type
• Some Financial Strategies Segments – less likely
• For each additional year of age over about 40: up to 5.5%
decline in probability of filing online
• Increasing take-up with income
SA316 online filing pilot | September 2009 | 27
Summary and conclusions (1)
• Pilot highly effective for some types of taxpayer – some
groups ten times more likely to adopt online filing
• Less impact for taxpayers with more complex affairs – such
taxpayers more likely to take on an agent who may often file
online on taxpayers behalf (irrespective of pilot/ control)
SA316 online filing pilot | September 2009 | 28
Summary and conclusions: future directions (2)
• Model could be used for future targeted exercises to
encourage online filing
• Potential benefit of lower costs, better overall customer
satisfaction and less risk of non-filing
• Potential to draw additional insights from application of
behavioural economics insights
• Discussion on possible operational applications in HMRC’s
work planned
SA316 online filing pilot | September 2009 | 29