Do Tax Administrative Interventions Targeted at Small Businesses Improve Tax Compliance? (WIDER Working Paper)

Summary

WIDER Working Paper No. 2021/17, by Maria Jouste, Milly I. Nalukwago and Ronald Waiswa for the United Nations University World Institute for Development Economics Research, evaluates two tax administration interventions in Uganda: a taxpayer register expansion and education programme and a new electronic filing system for presumptive tax. Using a difference-in-differences approach with administrative data on presumptive taxpayers and comparable small corporate income taxpayers, the authors find that the number of small business taxpayers filing returns and presumptive tax revenues rose substantially after the interventions. They argue the two interventions complement each other and find that the benefits of the register expansion programme outweigh its costs.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Full text

Jouste, Maria; Nalukwago, Milly I.; Waiswa, Ronald



Working Paper
Do tax administrative interventions targeted at small businesses
improve tax compliance and revenue collection? Evidence from
Ugandan administrative tax data


WIDER Working Paper, No. 2021/17


Provided in Cooperation with:
United Nations University (UNU), World Institute for Development Economics Research (WIDER)




Suggested Citation: Jouste, Maria; Nalukwago, Milly I.; Waiswa, Ronald (2021) : Do tax administrative
interventions targeted at small businesses improve tax compliance and revenue collection?
Evidence from Ugandan administrative tax data, WIDER Working Paper, No. 2021/17, ISBN
978-92-9256-951-8, The United Nations University World Institute for Development Economics
Research (UNU-WIDER), Helsinki,
https://doi.org/10.35188/UNU-WIDER/2021/951-8


This Version is available at:
https://hdl.handle.net/10419/229418

Standard-Nutzungsbedingungen:                                               Terms of use:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen             Documents in EconStor may be saved and copied for your personal
Zwecken und zum Privatgebrauch gespeichert und kopiert werden.              and scholarly purposes.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle            You are not to copy documents for public or commercial purposes, to
Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich        exhibit the documents publicly, to make them publicly available on the
machen, vertreiben oder anderweitig nutzen.                                 internet, or to distribute or otherwise use the documents in public.

Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen              If the documents have been made available under an Open Content
(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten,            Licence (especially Creative Commons Licences), you may exercise
gelten abweichend von diesen Nutzungsbedingungen die in der dort            further usage rights as specified in the indicated licence.
genannten Lizenz gewährten Nutzungsrechte.




                       https://creativecommons.org/licenses/by-nc-sa/3.0/igo/
WIDER Working Paper 2021/17



Do tax administrative interventions targeted at
small businesses improve tax compliance and
revenue collection?

Evidence from Ugandan administrative tax data


Maria Jouste,1 Milly I. Nalukwago,2 and Ronald Waiswa3




January 2021
Abstract: This paper conducts an impact evaluation of the effects of two tax administration
interventions—a taxpayer register expansion and education programme, and a new electronic filing
system for presumptive tax—on the number of small business taxpayers and presumptive tax
revenues in Uganda. Using a difference-in-differences approach and administrative data covering
both presumptive taxpayers and comparable small corporate income taxpayers, we find that the
number of small business taxpayers filing tax returns and presumptive tax revenues increased
substantially after the interventions. We argue that the interventions complement each other
because both interventions were established around the same years, and the taxpayer register
expansion programme focused on not only registering but also educating taxpayers with regard to
tax compliance. We analyse the cost-effectiveness of the taxpayer register expansion programme
and find that the benefits outweigh the costs.
Key words: tax administration, small businesses, tax compliance, electronic filing, impact
evaluation, administrative data
JEL classification: H25, H71, O17
Acknowledgements: This work is part of a larger research and capacity-building initiative
between the Uganda Revenue Authority (URA) and UNU-WIDER. Financing of the Finnish
Ministry of Foreign Affairs is gratefully acknowledged, as well as support to Jouste by the Kone
Foundation (grant no. 80-42327). This study includes information gathered during the authors’
personal discussions and correspondence with staff at URA. We thank colleagues at URA,
specifically Tina Kaidu, Nicholas Musoke, Dorothy Nakyambadde, and Joseph Okello, for their
comments and ongoing support of this work. We also thank Jukka Pirttilä, Pia Rattenhuber, and
Janne Tukiainen for their insightful feedback. For their comments, we thank participants in the
following: Helsinki Graduate School of Economics; Centre for the Study of African Economies
Conference 2019; International Institute of Public Finance Annual Congress 2019; Eighth
International Domestic Revenue Mobilization Workshop in Bonn; University of Uppsala;
University of Turku; Tax for Development Webinar Series 2020 at the Chr. Michelsen Institute.
The results and their interpretation presented here are solely the authors’ responsibility.




1   University of Turku, Turku and UNU-WIDER, Helsinki, Finland, corresponding author: jouste@wider.unu.edu
2 Uganda Revenue Authority, Kampala, Uganda 3 Uganda Revenue Authority, Kampala, Uganda

This study has been prepared within the UNU-WIDER project Building up efficient and fair tax systems—lessons based on
administrative tax data, which is part of the Domestic Revenue Mobilization programme. The programme is financed through
specific contributions by the Norwegian Agency for Development Cooperation (Norad).
Copyright © UNU-WIDER 2021 / Licensed under CC BY-NC-SA 3.0 IGO
Third-party content is not covered by the terms of the Creative Commons license. For permission to reuse third-party content,
please verify terms at the source and, if applicable, contact the rights holder.
Information and requests: publications@wider.unu.edu
ISSN 1798-7237 ISBN 978-92-9256-951-8
https://doi.org/10.35188/UNU-WIDER/2021/951-8
Typescript prepared by Merl Storr.
United Nations University World Institute for Development Economics Research provides economic analysis and policy advice
with the aim of promoting sustainable and equitable development. The Institute began operations in 1985 in Helsinki, Finland, as
the first research and training centre of the United Nations University. Today it is a unique blend of think tank, research institute,
and UN agency—providing a range of services from policy advice to governments as well as freely available original research.
The Institute is funded through income from an endowment fund with additional contributions to its work programme from
Finland, Sweden, and the United Kingdom as well as earmarked contributions for specific projects from a variety of donors.
Katajanokanlaituri 6 B, 00160 Helsinki, Finland
The views expressed in this paper are those of the author(s), and do not necessarily reflect the views of the Institute or the United
Nations University, nor the programme/project donors.
1     Introduction


In developing countries, weak institutions and low social norms of paying taxes create challenges
for the mobilization of domestic resources (Besley and Persson 2014). Besley and Persson (2014)
argue that broadening the tax base is more important for increasing tax revenues than changing
tax rates. On the other hand, Best et al. (2015) show that a simplified tax regime based on turnover
can lead to greater tax revenues compared with profit taxation in countries with a low tax capacity
and a large informal sector. In many low-income countries, governments have established a
presumptive tax regime for small and medium-sized enterprises where tax is based on a firm’s
turnover instead of profits, and tax rates are lower. The usual motivation for implementing a
presumptive tax is that taxpayers are not obliged to keep comprehensive records of their sales and
expenses, and hence tax liability is calculated from estimated turnover. In general, presumptive tax
collection is low, and it does not have fiscal importance for total tax revenues. For example, in
Uganda presumptive tax revenue is only 0.04 per cent of tax revenues (see Table 1). Nevertheless,
simpler tax regimes can be a first step to simplifying taxpaying, inducing formalization, and making
firms inclusive of social services.

Different tax regimes for small businesses are not enough on their own to reduce the informal
sector in developing countries. Tax administrative innovations and the adoption of new
technologies may help governments to improve tax compliance and reach a larger group of
business owners. For example, simpler online filing and payment of taxes may lower the costs of
paying and collecting taxes. However, technological innovations are not the only solutions to
increase tax compliance. Traditional methods, such as tax education campaigns or taxpayer register
programmes, are used to enhance taxpaying in developing countries. Several studies have evaluated
the impacts of formalization interventions such as business registration programmes and
information campaigns on firms’ formalization rate, and have found that in general such
interventions do not increase the number of formal firms (for literature reviews see e.g., Bruhn
and McKenzie 2014; Floridi et al. 2020). Most previous studies have examined formalization
interventions in Latin America and South-East Asia, and only a few have been conducted in Africa.
Benhassine et al. (2018) conducted a randomized experiment whereby they introduced an
‘entreprenant’ legal status in Benin; they found that the new legal status on its own did not affect
formalization, but when they made additional efforts, such as offering business training,
formalization increased by 16.3 percentage points. Lediga et al. (2020) examined the
synchronization of business and tax registers in South Africa, and they showed that although this
expanded the taxpayer register, firms failed to file and pay taxes after registration, which led to low
revenue gains.

In this study we analyse two different tax administrative interventions in Uganda: the Taxpayer
Register Expansion Project (TREP) and the new electronic filing (e-filing) system for presumptive
tax. The Uganda Revenue Authority (URA) implemented both interventions between 2013 and
2016. TREP started as a pilot programme in July 2013 in Kampala-based municipalities, and was
expanded twice, in July 2014 and July 2016, to cover more municipalities. However, TREP has not
yet reached nationwide coverage. The new e-filing system was introduced in July 2015 and is used
nationwide.

Both TREP and the new e-filing system are targeted at small businesses. TREP is mainly a taxpayer
register campaign, but it includes tax education as well. Moreover, TREP tries to harmonize the
administrative systems of different governmental institutions in Uganda, and as a result to make
business registration and the payment of taxes and other administrative fees easier. The key feature
of TREP has been the establishment of one-stop shops where businesses can deal with several

                                                  1
agencies in one visit. The new e-filing system for presumptive tax is an online tax form that small
businesses use to file their incomes. Previously, the presumptive tax form was an Excel form. The
change from Excel to the new e-filing form made filing simpler and more transparent. In the new
e-filing system, taxpayers only need to have access to URA’s website, where they fill in the form,
which automatically calculates the payable taxes and provides a payment receipt. The previous
Excel form was more complicated because it required taxpayers not only to have Excel on their
computers but also to know how to use it. After filling in the Excel form, they needed to submit
it to URA’s website, and then separately register the payment to get the receipt.

We analyse the impacts of these two interventions using several URA tax data sets: presumptive
and corporate income tax (CIT) returns, and taxpayer registration records. First, we examine the
impacts of the three TREP phases (I, II, and III) on the number of taxpayers, using the difference-
in-differences (DiD) method. We compare the number of presumptive taxpayers in geographical
areas where TREP I–III have been implemented with areas where TREP has never been active.
Further, we estimate the effects separately for the years before and after the establishment of one-
stop shops. Second, we estimate the effect of the new e-filing system on the number of taxpayers
using a similar DiD approach, but in this analysis we compare presumptive taxpayers with CIT
payers. In addition, we differently measure the impacts for the first and second years of the new
e-filing form. Third, we use the same DiD approach to analyse the impacts of the two interventions
on tax revenues. Finally, we analyse the costs and benefits of TREP. In summary, this study
evaluates whether the reforms have increased tax-filing compliance and made taxpaying easier, and
whether this has led to larger tax revenues while keeping the costs of the interventions reasonable.

The results of the analysis of TREP and the new e-filing system are as follows. First, we show that
the number of presumptive taxpayers increased after the interventions. Our preferred DiD
estimates of TREP (I–III) for the log number of taxpayers are positive and significant, varying
from 0.4 to 0.8. In particular, the findings suggest that later phases of TREP, i.e. after the
establishment of one-stop shops, had a large positive impact on the number of taxpayers, with an
approximately 70 per cent increase or more. We find that the new e-filing system substantially
increased the number of presumptive taxpayers. The increase was larger in the second year after
the reform than during the e-filing form’s implementation year—0.7 versus 1.5. The average
estimate for all years after the reform is around 1.1, and it indicates that the new e-filing system
approximately doubled the number of presumptive taxpayers.

Second, in the regressions for the new e-filing system, we find that the estimate of the interaction
effect of TREP and presumptive taxpayers is positive. Based on this positive estimate, we argue
that TREP and the new e-filing system have had complementary impacts. The mechanism behind
the findings is presumably that at one-stop shops, taxpayers are forced to register with URA before
they can apply for a trading licence, and during the same visit tax officers can educate and help
taxpayers to report their incomes to URA. The new e-filing form has made filing simpler, and
therefore tax officers can easily help taxpayers. In summary, the most feasible mechanisms are the
enforcement and provision of information at one-stop shops, and the lowering of compliance
costs. These mechanisms are described in the theory of tax compliance, which mostly emphasizes
that enforcement is a key factor to improve tax compliance, although other factors are also
mentioned and studied empirically (for a literature review, see Alm 2019). For instance, Alm (2019)
states that taxpayers’ knowledge about tax systems affects compliance, and laboratory experiments
have shown that administrative services that make taxpaying easier increase tax compliance. To
our knowledge, no previous studies have examined the impacts of improved tax administrative
services on tax compliance in developing countries.

Third, we find that both TREP and the new e-filing system raised tax revenues. Both interventions
approximately doubled presumptive tax collection. The increase in presumptive tax revenues was

                                                 2
larger when one-stop shops were established in TREP. Similarly, the impact of the new e-filing
system was larger in the second year after the reform. We also investigate the cost-effectiveness of
TREP and find that the average additional presumptive tax revenue collected from small
businesses after TREP was larger than TREP’s budget utilization.

Our paper contributes to the literature in several ways. First, we contribute to the increasing
literature that investigates tax compliance using administrative data in developing countries by
deepening the understanding of how large-scale taxpayer register programmes and simpler tax
forms affect tax compliance in low-income countries and whether these interventions increase
revenues in a cost-effective way. Previous studies that have investigated tax compliance using
administrative data and quasi-experimental settings have mostly been from middle-income
countries such as Pakistan and South Africa. For instance, using Pakistani administrative tax data,
Waseem (2018, 2020) analyses how changes in tax rates and withholding mechanisms of value
added tax (VAT) have impacted on tax compliance. Lediga et al. (2020) evaluate administrative
interventions that synchronize business and tax registers, employing South African tax records.
However, most previous studies that have used administrative data to examine tax compliance
have been randomized controlled trials. For example, Mascagni et al. (2019) analyse a tax education
experiment in Rwanda, and Carillo et al. (2017) examine the effectiveness of third-party
information using an experimental setting in Ecuador. Furthermore, several studies have
investigated the impacts of tax information letters (e.g., a novel study by Brockmeyer et al. (2019),
and a literature review by Mascagni (2018)). Thus, our paper is one of the first studies to use
administrative tax data and a natural experimental setting to evaluate the impact of administrative
interventions in a low-income country.

Second, our paper adds to the literature on formalization interventions in developing countries,
which includes studies in Latin America and South-East Asia reviewed by Bruhn and McKenzie
(2014) and Floridi et al. (2020), and novel research in Africa by Lediga et al. (2020) and Benhassine
et al. (2018). In particular, we extend the knowledge about how formalization interventions can
impact on not only firm registration but also tax-filing compliance, and whether this can be
achieved cost-effectively.

Finally, we contribute to the scarce literature evaluating electronic filing in developing countries
by examining the impacts of the simplified e-filing form in Uganda. Only a few studies have
investigated the impacts of electronic filing. For instance, using survey data, Yilmaz and Coolidge
(2013) evaluate the compliance costs of e-filing by small and medium-sized firms in South Africa,
Ukraine, and Nepal; they conclude that the compliance costs depend on the e-filing system and
the compulsory or obligatory nature of the e-filing. Another example is the study conducted by
Okunogbe and Pouliquen (2018). They analyse how e-filing affects small and medium-sized firms’
compliance costs, tax payments, and bribe payments in Tajikistan, using an experimental setting
and both survey and administrative tax data. Their results show that e-filing reduces not only
compliance costs but also tax and bribe payments.

The rest of the paper is organized as follows. Section 2 provides the institutional background and
explains the reforms. Section 3 describes the data and presents descriptive evidence. Section 4
presents the chosen empirical strategy. Section 5 reports the main findings and discusses the cost-
effectiveness of TREP. Finally, section 6 concludes.




                                                 3
2     Institutions


Uganda is a low-income country in sub-Saharan Africa: its gross domestic product (GDP) per
capita in 2017 was US$2,075 purchasing power parity (World Bank 2018). It is a member of the
East African Community. Tax revenue collection is low in Uganda compared with other East
African Community member countries. Figure 1 shows the development of tax revenues in
Uganda, Kenya, Rwanda, and Tanzania. The share of tax revenues increases after 2010 in Uganda,
but it is still lower than in Kenya, Rwanda, or Tanzania. For example, in 2015 tax revenue was
10.45 per cent of GDP in Uganda, 16.26 per cent of GDP in Kenya, 15.71 per cent of GDP in
Rwanda, and 10.49 per cent of GDP in Tanzania (UNU-WIDER 2020). Moreover, in 2015 the
average tax revenue in sub-Saharan African countries was 15.75 per cent of GDP, which was
higher than in Uganda.

As in many developing countries, the informal sector is large in Uganda. The Uganda Bureau of
Statistics estimates that approximately 87 per cent of total employment is informal, and in the fiscal
year 2017–18, the size of the informal sector was 51.2 per cent of GDP (Uganda Bureau of
Statistics 2016, 2019). A World Bank report in 2017 studied the informal sector in the Greater
Kampala area (World Bank 2017). 1 It reported that 57 per cent of firms in the Greater Kampala
area were informal, and more than half of them had annual sales below the lowest tax threshold.
This highlights that most of the informal firms in Kampala are microbusinesses.
Figure 1: Tax revenues in per cent of GDP for Uganda, Kenya, Rwanda, and Tanzania




Source: authors’ visualization based on data from UNU-WIDER (2020).

In Uganda, citizens’ access to the Internet and capacity to use information and communications
technology are limited. The adult literacy rate was 76.5 per cent in 2018 (UNESCO Institute for
Statistics 2020). Only 23.71 per cent of the total population use the Internet in Uganda, and there
were 57.27 mobile phone subscriptions per 100 people in 2018 (International Telecommunication


1
  The report was based on the National Manpower Survey collected by the Uganda Bureau of Statistics. For more
information about this survey, see World Bank (2020b).

                                                     4
Union 2020). Uganda’s National Information Technology Survey for 2017–18 found that 70.9 per
cent of individuals had a mobile phone, but only 15.8 per cent of those had a smartphone (National
Information Technology Authority Uganda 2018). Moreover, 10.8 per cent of all households
participating in the survey had at least one person who could access the Internet at home. The
survey also asked questions related to e-governance: among all individuals, 96.6 per cent said that
they interacted face to face with government officials, while only 2.3 per cent used institutional
websites. In e-governance services, 62.6 per cent of all individuals had used mobile payments to
pay utility bills, 49.9 per cent had used online registration to receive their taxpayer identification
number (TIN), and 35.4 per cent had paid taxes online, but only 13 per cent had filed their tax
returns online in the previous 12 months.

2.1      Domestic tax laws of Uganda

The domestic tax laws of Uganda consists of several tax acts, 2 among which we are interested in
the Income Tax Act (of 1 July 1997, Cap 340 of the Laws of Uganda 2000) and the Value Added
Tax Act (of 1 July 1996, Cap 349 of the Laws of Uganda 2000). The Income Tax Act includes the
taxation of individuals, partnerships and partners, trusts and beneficiaries, and companies and
shareholders. It provides tax rates for small business taxpayers (referred to as presumptive tax),
individuals, companies (referred to as corporate tax), trustees and retirement funds, dividends,
rents, and other special sources of income. The Value Added Tax Act contains rules about VAT.

According to the Income Tax Act, businesses can pay CIT, presumptive tax, or personal income
tax, depending on the size and type of business. The tax rules are different for CIT, presumptive
tax, and personal income tax. First, CIT is paid by companies 3; for other than mining companies
the tax rate is 30 per cent of taxable income, which is based on profits; this tax rate has been same
since 1997. Second, presumptive tax can apply to either an individual business or a company, and
the tax schedule consists of several tax rates and income thresholds (see Tables A1 and A2 in
Appendix A for more details). Since 1 July 2015, businesses have been eligible to pay presumptive
tax if their estimated turnover is over UGX10 million but does not exceed UGX150 million. The
estimated turnover is the taxable income for presumptive taxpayers, because they are not eligible
for any tax deductions or exemptions. Third, personal income tax is paid by individual businesses,
and they follow personal income tax rules. 4 Taxpayers can choose between the presumptive, CIT,
and personal income tax schedules, even if their turnover is between the lower and upper
thresholds of the presumptive tax schedule, but in this case they need to notify the URA
commissioner in writing (URA 2015). 5 If a taxpayer chooses to pay CIT or personal income tax
instead of presumptive tax, they need to keep more detailed accounts, but they also have the right
to apply tax deductions and exemptions.

The Value Added Tax Act determines VAT rules. The VAT rate is 18 per cent and has been
unchanged since 2006 (Value Added Tax (Rate of Tax) Order 2006, of 1 July 2005). However,
there have been changes in zero-rated items and the threshold value, which determines the


2
 The domestic tax acts are the Income Tax Act, the Value Added Tax Act, the Excise Duty Act, the Tax Procedures
Code Act, the Stamp Duty Rates, the Lotteries and Gaming Act, the Tax Appeals Tribunals Act, and the Finance Act.
3
    A company is a body of persons, corporate or unincorporated, or a unit trust (Income Tax Act, of 1 July 1997).
4
 Personal income taxation is not the interest of this paper, and therefore we do not provide detailed information
about these tax rules. For more information, see the Income Tax Act (of 1 July 1997, Cap 340 of the Laws of Uganda
2000).
5
  In general, taxpayers choose their tax regime based on their business type and whether or not they keep
comprehensive accounts.

                                                           5
obligation to pay VAT (Value Added Tax (Amendment) Act 2015, of 1 July). The zero-tax
threshold increased on 1 July 2015. Before the change, firms were exempted from paying VAT if
their annual sales were below UGX50 million. After 1 July 2015, the annual sales threshold rose
to UGX150 million. This means that presumptive taxpayers have been always exempted from
paying VAT.

In Table 1, we present the share of tax revenues by different tax types. Domestic taxes constitute
around 55 per cent of gross revenue. For example, in 2016–17 6 the largest share of domestic taxes
is collected from direct taxes (32.42 per cent), of which pay-as-you-earn 7 tax covers over half. VAT
has the second largest share of gross revenue (22.04 per cent) among the different tax types.
Presumptive tax is by far the smallest part of tax revenues, covering only 0.04 per cent of gross
revenue. Overall, the share of tax revenues by different tax types remains fairly similar from 2012–
13 to 2016–17.
Table 1: Percentage shares of tax revenue by different tax types, 2012–13 to 2016–17

                                           2012-13          2013-14         2014-15         2015-16         2016-17
Domestic taxes                              57.08            55.76           54.90           54.82           56.02
Direct taxes                                32.49            31.33           32.11           32.24           32.42
CIT                                          7.98             5.81            7.07            6.37            5.93
Presumptive tax                                                                               0.01            0.04
PAYE                                         15.98             16.68         15.95           15.69           16.40
Indirect taxes                               23.11             22.68         21.24           21.25           22.04
VAT                                          17.08             16.16         14.93           15.41           15.69
Excise duty                                   6.03              6.52         6.31             5.84            6.36

Note: percentage shares are calculated using gross revenue. Presumptive tax is reported separately only after
2015–16, when the new e-filing form came into effect.
Source: authors’ compilation based on revenue statistics from URA.

2.2      Tax administrative reforms

The Ugandan government implemented several tax and administrative reforms targeted at small
businesses between 2009 and 2018. Figure 2 summarizes the reforms and their timelines. The first
administrative reform was the implementation of the e-tax system, which started in 2009; by 2012,
e-tax covered all tax offices in Uganda. The second administrative reform was the three phases of
TREP, starting in July 2013. Finally, in July 2015, the new e-filing system was introduced for
presumptive taxpayers. The e-tax and e-filing systems and TREP are explained in more detail in
subsequent sections.

During the period when the tax administrative reforms came into effect, the Ugandan government
changed the presumptive tax rate twice. The first presumptive tax reform, in July 2014, increased
the tax rate from one to three per cent of estimated sales (Income Tax (Amendment) Act 2014, of
1 July). The second reform was implemented the next year. It was larger than the 2014 reform,
because both the tax rate and the thresholds of the tax brackets changed (Income Tax
(Amendment) Act 2015, of 1 July). The 2015 reform substantially increased the highest eligibility


6 The fiscal year in Uganda runs from 1 July to 30 June. In this study, we refer to fiscal years using the convention

‘2012–13’ for a fiscal year starting on 1 July 2012 and ending on 30 June 2013. Taxpayers can apply a substituted fiscal
year—for example, a calendar year—if it is more convenient for the company’s accounting. In these cases, the fiscal
year for tax purposes is based on the end date of the tax return period. For instance, if a firm’s tax return period ends
on 31 December 2014, the fiscal year is referred to as 2014–15.
7
    Pay-as-you-earn is tax on employment income: employers remit tax on behalf of employees and pay it to URA.

                                                           6
threshold for presumptive tax, from UGX50 million to UGX150 million. At the same time, the
tax rate dropped from three to 1.5 per cent. After the 2015 reform, there were minor changes in
the tax schedule (Income Tax (Amendment) Act 2016, of 1 July). In this paper, we do not focus
on the tax reforms; therefore, all the details of the tax changes are provided in Tables A1 and A2
in Appendix A.
Figure 2: Timeline of different reforms targeted at small businesses and presumptive taxpayers, July 2009 to July
2018




Note: the e-tax implementation period was the time when e-tax did not operate in all tax offices. The e-tax
affected all tax types. E-filing, online form: new e-filing system for presumptive tax. Presumptive tax increase:
increase of the tax rate and zero-tax threshold. Presumptive tax decrease: decrease of the tax rate and changes
in tax bracket thresholds. For more information, see Tables A1 and A2 in Appendix A. Minor tax changes:
increase in tax rate on pharmacies from UGX100,000 to UGX250,000, and exclusion of clinics from the
presumptive tax regime. VAT threshold increase: registration threshold for VAT increased from UGX50 million to
UGX150 million.
Source: authors’ visualization based on internal URA documents, the Income Tax Act (of 1 July 1997, including
Income Tax (Amendment) Acts of 2014, 2015, 2016, all of 1 July), and the Value Added Tax (Amendment) Act
2015 (of 1 July).

E-tax system and filing tax returns: the e-filing reform of July 2015

URA established the e-tax system in 2009. It was first introduced as a pilot programme in three
tax offices in 2009, from where it was gradually rolled out to other tax offices. By the end of the
fiscal year 2011–12, the e-tax system was in place in every tax office in Uganda. The e-tax system
is an online platform where citizens can report their incomes for tax purposes, pay taxes, and
register as taxpayers. On registration, a taxpayer receives the TIN that is used in the e-tax system.

Taxpayers are subject to filing their tax returns using the e-tax system. In general, the filing is done
using downloadable Excel forms from the URA website. After downloading the Excel form, a




                                                        7
taxpayer8 completes the return form, which is different for different types of taxpayers. For
instance, a CIT form includes more spreadsheets to complete than a presumptive tax form,
because CIT payers are obliged to keep comprehensive accounts. When the return is successfully
completed, the taxpayer submits it to URA’s e-tax system.

URA redesigned the e-tax system for presumptive taxpayers, and in July 2015 it introduced a new
e-filing system for presumptive tax returns. The new e-filing system does not include an Excel
form; instead, taxpayers can directly file their tax returns using an online form on URA’s e-tax
website. Before July 2015, presumptive tax returns were filed using the Excel form described
above. The new e-filing form is only for presumptive tax returns; all other tax returns are filed
using Excel forms.

The new e-filing system for presumptive taxpayers also contains other features besides the
completion of returns on the online form. First, the online form includes payment instructions at
the end of the filing. The system generates a receipt that taxpayers can print and submit to the
bank, or they can use other payment methods such as mobile money or credit cards. Second,
taxpayers can use the same e-filing system if they want to amend their submitted tax return.
However, if a taxpayer wants to decrease the annual turnover submitted in an earlier return in the
same year, they need to visit their tax office in person. Finally, one helpful feature is that when
taxpayers insert their estimated annual sales into the e-filing form, it automatically calculates the
payable taxes. Taxpayers can observe how their payable taxes change if they change their incomes
while they fill in the form.

TREP

The Ugandan government launched TREP in July 2013 to encourage citizens, particularly small
businesses, to register as taxpayers and receive a TIN. TREP is jointly organized by URA, the
Uganda Registration Services Bureau (URSB), Kampala Capital City Authority, and local
governments. The objectives of TREP are to formalize businesses, educate them about taxes,
reduce compliance costs, and harmonize different government agencies’ revenue and tax
administration systems (URSB 2017).

The first phase of TREP (TREP I) started in July 2013 (fiscal year 2013–14) in Kampala-based
municipalities (see Table 2). The second phase, TREP II, followed in the fiscal year 2014–15 in
municipalities located in the Wakiso district. The last phase, TREP III, expanded the project to 30
municipalities outside the Kampala and Wakiso districts in July 2016. TREP III did not lead to
coverage of the whole country, and in 2018 there were municipalities that were not included in the
project.

In TREP, various working methods have been used (described in Appendix B) to boost the
registration of taxpayers. In general, the timeline for TREP was as follows: (1) engage local
councils, leaders, and stakeholders in the municipalities; (2) make an agreement that TREP can
start; (3) decide which TREP methods to use in the municipalities. In the first two years, the main
working method was door-to-door visits. Door-to-door visits continued in all three TREP phases,
but new methods were introduced as well (see Appendix B). In 2015–16, the working methods
expanded to the establishment of ten one-stop shops in Kampala. At a one-stop shop, business
owners can register their businesses with different authorities, become taxpayers, and even file and
pay their taxes, all in one visit. In 2015–16, the first public awareness campaigns—for instance,


8
 Under the Income Tax Act (of 1 July 1997, Cap 340 of the Laws of Uganda 2000), a taxpayer is any person who
obtains income subject to tax. A taxpayer may be, for instance, a company, an individual, or a trustee.

                                                     8
client sensitization workshops and radio talk shows—were launched (see Table 3). In 2016–17,
one-stop shops were established in all the other 34 TREP municipalities outside Kampala, and
more types of public awareness campaigns were used to reach a larger group of taxpayers (see
Tables 2 and 3).

TREP has been mainly targeted at small businesses, because the level of compliance among these
businesses is very low and URA has faced challenges to increase tax revenue collection from them.
However, TREP may have had some impact on medium-sized and larger businesses and individual
businesses, because some working methods—e.g., radio talk shows and newspaper strips—are
visible to everyone.
Table 2: TREP implementation: start dates and municipalities

    TREP I:          TREP II:                                        TREP III:
  operational      operational                              operational from 1 July 2016
  from 1 July      from 1 July
     2013              2014
   Kampala-       Municipalities                    Municipalities outside Kampala and Wakiso
     based          in Wakiso
 municipalities       district
  1 Makindye        1 Entebbe         1 Arua         7 Ishaka-      13 Kasese        19 Lugazi     25 Tororo
                                                     Bushenyi
   2 Central       2 Makindye        2 Busia          8 Jinja        14 Kisoro      20 Masindi    26 Mbarara
                   Ssabagabo
   3 Nakawa          3 Kira        3 Fort Portal     9 Kabale       15 Kitgumu       21 Mbale      27 Mityana
  4 Kawempe        4 Nansana          4 Gulu         10 Kamuli      16 Koboko      22 Ntungamo     28 Moroto
   5 Rubaga                          5 Hoima       11 Kapchorwa      17 Kumi       23 Rukungiri   29 Mubende
                                     6 Iganga       12 Masaka         18 Lira        24 Soroti     30 Mukono

Source: authors’ compilation based on documents from URA.



Table 3: Summary of TREP awareness campaigns

                                                                       2015-16         2016-17      2017-18
Stakeholder engagements                                                   27             77           104
Quarterly messages to staff                                                1              7            0
Client sensitization workshops/tax clinics                               116             35           151
Public notices                                                             0             17            8
Newspaper strips                                                           0             24           14
Barazas                                                                    0              1            0
Radio advertisements                                                       0            4,677        7,700
Public van communications                                                  0             36            7
Radio talk shows                                                          30             24           30
TV scripts                                                                 0             60           105
Press briefings                                                            1              1            0

Note: the column ‘2017–18’ shows campaigns as at the end of December 2017. See Appendix B for more
information about the nature of the different awareness campaigns.
Source: authors’ compilation based on documents from URA.




                                                        9
3        Data and descriptive evidence


We use administrative tax data from URA’s e-tax system. The data includes presumptive tax
returns from July 2009 to June 2018, and CIT returns 9 from July 2009 to June 2017. The
presumptive tax returns data consists of two separate data sets, because in July 2015 the new e-
filing form changed how data was generated in URA’s database. In addition to the returns data,
we add an address variable for CIT payers from the non-individual TIN registration form. 10

These data sets are appended and merged to create the final harmonized data set for the analysis.
We exclude a minor number of outliers from our analysis: for example, we drop observations that
have negative taxes or where the presumptive taxpayer’s turnover is above the highest threshold
for presumptive tax (either UGX50 million or UGX150 million, depending on the fiscal year).
Next, we exclude a small number of CIT return observations that do not have a clear address
variable, because our estimations require information about firms’ geographical locations. Finally,
we drop taxpayers who have shifted from presumptive tax to the CIT regime or vice versa, because
we want to keep the treatment (presumptive) and control (CIT payers) groups clean and
uncompromised by such shifts. 11

Table 4 presents summary statistics on the main outcome variables from presumptive and CIT
returns data between 2012–13 and 2017–18. The presumptive taxpayers are on average small
businesses with a turnover under UGX20 million and payable tax under UGX250,000. The CIT
payers have larger average turnovers and payable tax than the presumptive taxpayers. Naturally,
the large difference in payable tax comes from the different tax rules. The CIT data that we use
covers only taxpayers with a turnover less than or equal to UGX400 million.
Table 4: Summary statistics from presumptive tax and CIT returns data, 2012–13 to 2017–18

Year                  Average turnover                  Average payable tax            Number of taxpayers (total)
                 Presumptive         CIT            Presumptive        CIT            Presumptive        CIT
2012-13          13,709,442      68,029,973           118,765       3,225,070             549           11,372
2013-14          16,617,642      64,120,527           148,750       2,190,842            1,868          14,779
2014-15          11,282,286      62,647,449           257,547       2,086,218            2,175          17,160
2015-16           6,731,984      57,042,328           212,467       2,163,511            5,716          20,364
2016-17          16,864,458      53,655,654           223,858       1,948,385           24,738          21,260
2017-18          19,310,028                           220,399                           38,367

Note: statistics on presumptive taxpayers are generated using all observations in the presumptive tax returns
data. Statistics on CIT payers are from the data set that covers only firms with turnover less than or equal to
UGX400 million. All monetary values in UGX.
Source: authors’ calculations based on URA administrative tax returns data.

Table 4 shows a rapid increase in the number of presumptive taxpayers between 2015 and 2016.
In the fiscal year 2016–17, the presumptive tax returns data has over 24,000 observations. By
contrast, the number of CIT payers grows more constantly over the years. Figure 3 visualizes the
clear difference in growth of the numbers of observations in the two data sets.



9
 We refer to CIT returns when the returns are extracted from income tax return forms for non-individuals in URA’s
e-tax system.
10
     For example, companies apply for TINs using the non-individual TIN registration form.
11
     The dropped firms represent around 4.5 per cent of total observations in the data.

                                                           10
Figure 3: Number of taxpayers from presumptive tax and CIT returns data, 2011–12 to 2017–18




Note: the left panel presents the number of presumptive taxpayers, and the right panel presents the number of
CIT payers with turnover less than UGX400 million.
Source: authors’ calculations based on URA administrative tax returns data.



Figure 4: Number of taxpayers from presumptive tax and CIT returns data in the analysis of the new e-filing
system




Note: figure includes presumptive taxpayers with turnover less than UGX50 million, and CIT payers with turnover
between UGX150 million and UGX400 million. The dashed line indicates July 2015, when the new e-filing form
came into force. The figure is descriptive and not intended to be used to analyse the significance of pre-reform
time trends (cf. Figure 7).
Source: authors’ calculations based on URA administrative tax returns data.

To evaluate the new e-filing system, we further restrict the data to cover only presumptive
taxpayers whose turnover is UGX50 million or less, and CIT payers with turnover between
UGX150 million and UGX400 million. In the next section we explain our chosen methodology
in detail. The number of taxpayers shown in Figure 3 is larger than in Figure 4 because Figure 4
shows only the trend in the number of taxpayers using the restricted data. Thus, Figure 4 visualizes
more clearly that one year after the introduction of the new e-filing form, presumptive taxpayers
with turnover of UGX50 million or less have a massive growth in numbers compared with CIT
payers with turnover of UGX150–400 million. Similarly, tax revenues from these presumptive

                                                       11
taxpayers grow more than tax revenues from CIT payers, as seen in Figure 5. We normalize the
value of tax revenue in 2014–15 to one. This allows us to compare the trend in tax revenues instead
of showing different levels of tax revenues.
Figure 5: Tax revenues from presumptive tax and CIT returns data in the analysis of the new e-filing system




Note: tax revenues are normalized to value one in year 2014–15 to visualize the development of tax revenues.
The figure includes presumptive taxpayers with turnover less than UGX50 million, and CIT payers with turnover
between UGX150 million and UGX400 million. The dashed line indicates the reform time of July 2015.
Source: authors’ calculations based on URA administrative tax returns data.



4      Empirical strategy


In this section, we explain the research design that we use to evaluate the impacts of the
administrative interventions—i.e. TREP and the new e-filing system—on the number of taxpayers
and tax revenues. First, we estimate the effects of TREP using the DiD approach and employing
area-based variation, which comes from the municipal-level TREP implementation schedule.
Second, we concentrate on the new e-filing system and its impacts by using a similar DiD
approach. Because the new e-filing system affects all presumptive taxpayers countrywide, we use
CIT payers as a control group. Finally, we describe how we use the same DiD methods to analyse
the impacts of TREP and the new e-filing system on tax revenues.

4.1    Estimation of TREP

We conduct a DiD analysis to estimate the impact of TREP on the number of taxpayers. In this
paper, we focus on TREP’s effects on presumptive taxpayers, since TREP is mainly targeted at
small informal businesses, for which the presumptive tax schedule is designed. Therefore, we only
use presumptive tax returns data in this analysis.

TREP was implemented in phases in different municipalities (see Table 2), and it did not reach
countrywide coverage. Thus, the phased implementation of TREP creates a natural experimental
setting where some municipalities are treated by TREP and some are not. To compare the numbers


                                                       12
of taxpayers in the treated and control municipalities, we aggregate the individual-level
presumptive tax returns data to industry*geographical area cells where a geographical unit is a
county, 12 which includes municipalities, and an industry unit is a one-digit International Standard
Industrial Classification (ISIC) code. The outcome variable is the number of taxpayers in the
industry*geographical area cell in a year.

Because we are investigating how the number of presumptive taxpayers changed due to TREP, we
restrict our data to presumptive taxpayers who were able to pay presumptive tax between 2012–
13 and 2017–18. This leaves us with taxpayers whose turnover was less than UGX50 million,
which was the highest threshold for eligibility to pay presumptive tax before July 2015. In addition,
this restriction allows us to control for the presumptive tax reforms, since the tax reforms similarly
affected this group of presumptive taxpayers.

We analyse the three phases of TREP separately because TREP started in different years in
different municipalities. For this reason, we have three treatment groups: (1) municipalities where
TREP I is active; (2) municipalities where TREP II is active; (3) municipalities where TREP III is
active. Our control group is the rest of the counties, where TREP has never been active.

We use data from 2012–13 to 2017–18 because the e-tax was fully implemented by the end of the
fiscal year 2011–12, and 2017–18 is the last year for which we have complete data. As a result, we
have only one pre-reform year to use for the analysis of TREP I (the first TREP implementation
phase), while TREP II has two pre-reform years, and TREP III has four.

The estimated DiD equation is:

       𝐿𝐿𝐿𝐿𝐿𝐿 𝑁𝑁𝑐𝑐,𝑡𝑡 = 𝛽𝛽0 + ∑ 𝛽𝛽1,𝑐𝑐 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐 + ∑ 𝛽𝛽2,𝑡𝑡 𝑌𝑌𝑌𝑌𝑌𝑌𝑟𝑟𝑡𝑡 + 𝛽𝛽3 (𝑇𝑇𝑇𝑇𝑇𝑇𝑃𝑃𝑐𝑐 𝑥𝑥𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 ) + 𝜀𝜀𝑐𝑐,𝑡𝑡   [1]

𝑁𝑁𝑐𝑐,𝑡𝑡 is the number of taxpayers in cell c (c=industry*geographical area) in year t. 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐 is a dummy
variable to control for permanent differences between industry*geographical areas. 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌t are year
controls. 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑐𝑐 is one if a geographical area in cell c is a municipality where TREP I–III is active,
and zero if a county belongs to the control group. The interaction variable 𝑇𝑇𝑇𝑇𝑇𝑇𝑃𝑃𝑐𝑐 𝑥𝑥𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 takes
value one if cell c is treated by TREP I–III and the observation is after the treatment year, which
is 2013–14 for TREP I, 2014–15 for TREP II, and 2016–17 for TREP III. The coefficient 𝛽𝛽3 is
the variable of interest.

In addition to the analysis of the overall effect of different TREP phases, we divide the post-
reform period into two parts to capture the effects of different treatment methods in TREP I–III.
The first part includes years when there were no one-stop shops in place; the second part contains
years after the establishment of the one-stop shops. For TREP III we have only one 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡
variable, because one-stop shops were set up at the same time as TREP started in the
municipalities. Thus, different 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 variables identify the impacts of different treatments of
TREP.

To identify the impact of TREP, the treated and control municipalities need to have parallel trends
in the number of taxpayers in the absence of the treatment. We test this hypothesis indirectly by
checking whether the parallel trends were the same in the period before TREP started. Another


12
  In Uganda, geographical areas comprise regions, districts, counties, subcounties, and parishes, in descending order
based on the size of the administrative unit. County-level units are municipalities, which are mostly urban centres (e.g.,
Mbarara municipal council), and other, more rural areas, including smaller towns (e.g., Serere county).

                                                                  13
concern regarding identification is that e-filing and tax reforms were introduced at the same time
as TREP, which might affect the outcome. However, the implementation of the new e-filing
system and tax reforms covers all presumptive taxpayers, and therefore it should affect the
treatment and control groups similarly. To the best of our knowledge, there were no other reforms
or shocks during 2012–18 besides those already mentioned. Furthermore, we assume that no
spillovers occurred between the treated and control municipalities.

4.2       Estimation of the new e-filing system

The new e-filing system replaced the old Excel-based e-filing form for presumptive tax in July
2015, and it affected all presumptive taxpayers similarly. Therefore, we cannot use area-based
treatment and control groups to estimate the impact of the new e-filing system on the number of
taxpayers, as we do in the analysis of TREP. Instead, we employ firms that pay CIT as a control
group. From the CIT returns data we include only firms with a turnover between UGX150 million
and UGX400 million in our analysis, because these firms have never been eligible to pay
presumptive tax. Because the new e-filing only impacted on presumptive returns, we use the
presumptive tax returns data to create a treatment group. We restrict the presumptive tax returns
data to taxpayers with a turnover under UGX50 million, because they were eligible to pay
presumptive tax in the estimation period 2012–13 to 2016–17.

We aggregate taxpayer-level data to industry*geographical area level using districts and TREP areas
(shown in Table 2) as geographical area variables, and one-digit ISIC codes as industry variables.
The best option would have been to use a county as a geographical unit, as in the analysis of TREP,
but we do not have county-level information in the CIT returns data. The only information that is
in the data is on districts and municipalities. Because TREP was implemented in municipalities,
the municipal variable in the CIT returns data is needed to create a TREP area variable in order to
control for the TREP that was implemented during our estimation period. Moreover, to compare
the numbers of taxpayers in the presumptive and CIT returns data, we further aggregate the data
to industry*geographical area*(presumptive or CIT) cells. In this way, we create an outcome
variable: the number of taxpayers in a cell in a year.

We use two DiD equations. The first equation includes a simple control variable if TREP I–III is
active; the second equation extends the first by adding the interaction term for the active TREP
and a treatment group (i.e. presumptive). The estimated equations are:

          𝐿𝐿𝐿𝐿𝐿𝐿 𝑁𝑁𝑐𝑐,𝑡𝑡 = 𝛽𝛽0 + ∑ 𝛽𝛽1,𝑐𝑐 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐 + ∑ 𝛽𝛽2,𝑡𝑡 𝑌𝑌𝑌𝑌𝑌𝑌𝑟𝑟𝑡𝑡 + 𝛽𝛽3 (𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃c 𝑥𝑥𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 ) +
          𝛽𝛽4 (𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 𝑥𝑥𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌2014/15 ) + 𝛽𝛽5 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 + 𝜀𝜀𝑐𝑐,𝑡𝑡                         [2]

          𝐿𝐿𝐿𝐿𝐿𝐿 𝑁𝑁𝑐𝑐,𝑡𝑡 = 𝛽𝛽0 + ∑ 𝛽𝛽1,𝑐𝑐 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐 + ∑ 𝛽𝛽2,𝑡𝑡 𝑌𝑌𝑌𝑌𝑌𝑌𝑟𝑟𝑡𝑡 + 𝛽𝛽3 (𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃c 𝑥𝑥𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 ) +
          𝛽𝛽4 (𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 𝑥𝑥𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌2014/15 ) + 𝛽𝛽5 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 +
          𝛽𝛽6 (𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 𝑥𝑥𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐 ) + 𝜀𝜀𝑐𝑐,𝑡𝑡                                               ]3]

𝑁𝑁𝑐𝑐,𝑡𝑡 is the number of taxpayers in cell c (c=industry*geographical area*(presumptive or CIT)) in
year t. 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐 are time-invariant fixed effects for industry*geographical area*(presumptive or CIT)
cells. 𝑌𝑌𝑌𝑌𝑌𝑌𝑟𝑟t are dummies for each fiscal year. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is one if cell c is
industry*geographical area*presumptive, and zero if it is industry*geographical area*CIT. 13 The
interaction variable 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 𝑥𝑥𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 takes value one if cell c is from the presumptive data
and the year is after the introduction of the new e-filing system (i.e. July 2015).


13
     The main effect of 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is not needed, because the cell dummies include it.

                                                                    14
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 𝑥𝑥𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌2014/15 controls for the presumptive tax reform. 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 is a time-
varying variable to capture the effect of TREP; it takes value one if cell c consists of a geographical
area where TREP is implemented and year 𝑡𝑡 is after TREP is active, and it is zero when either a
geographical area is not impacted on by TREP or the year is before TREP started. 14 In Equation
[3], we add the interaction term 𝑇𝑇𝑇𝑇𝐸𝐸𝑃𝑃_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 𝑥𝑥𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 , which is one if cell c is
industry*geographical area*presumptive and 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 is one. This interaction captures the
complementary effects of TREP and the new e-filing system on the number of taxpayers. In both
equations, the variable of interest is 𝛽𝛽3, the interaction term of being treated by the new e-filing
system and the post-reform period. In addition to the basic analysis, we investigate the timing of
the new e-filing system by estimating the first and second years after the reform separately.

The underlying assumption for the DiD approach is that the treatment and control groups would
have similar trends over time in the absence of the treatment. We test this assumption indirectly
by estimating the treatment effect for the pre-reform period, and we visually show whether the
trend is similar for the treatment and control groups. Furthermore, the control group needs to be
independent of the reform, i.e. not affected by the new e-filing system. In general, a firm can switch
between the presumptive and CIT regimes. Therefore, we drop taxpayers who shifted between the
CIT and presumptive tax regimes from the data before aggregating the data in cells. As a further
robustness check, we test whether the specification of the control group changes the results. 15

One possible problem in the identification is that other reforms happened at the same time as the
new e-filing system came into effect. In 2012–17 the presumptive tax schedule was changed twice,
and from 2013 onwards TREP was rolled out. First, we control for the presumptive tax rate change
by using the presumptive tax rate change in 2014–15 as a proxy for the change in 2015–16. Without
this control, the interaction term 𝛽𝛽3 would capture the impacts of both the change of the e-filing
system and the decline in tax rates because the two reforms came into force at the same time in
July 2015. Second, we control for the effect of TREP. Although TREP is mainly designed for
small businesses that pay presumptive tax, it might impact on CIT payers, because advertising
campaigns such as radio talk shows are publicly broadcasted. Furthermore, small corporations
might also have visited the one-stop shops. As a result, we control for TREP for both the
treatment and control groups in equation [2].

4.3    Estimation of tax revenues

We evaluate the impact of TREP and the new e-filing system on tax revenues using the same DiD
approach as described in Sections 4.1 and 4.2. The outcome variable in the regression is payable
taxes, rather than the number of taxpayers.

Obviously, the presumptive tax reforms affect revenue calculations. For example, the tax reform
in 2015–16 decreased the presumptive tax rate, but it also changed the tax bracket thresholds (see
Tables A1 and A2). Both changes have direct effects on the tax revenues collected from
presumptive taxpayers. As a robustness check, we calculate payable presumptive taxes from
turnover using a constant presumptive tax rate for all years. This controls for tax reforms and
shows how tax revenues develop due to changes in the number of taxpayers.



14
  For example, 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 is one when cell c has the Wakiso district and Entebbe municipality as a geographical
area and year t is 2016–17.
15
   For example, we use CIT returns data that includes all firms with turnover under UGX400 million before
aggregating the data at the cell level.

                                                             15
5     Results


5.1   Number of taxpayers

In this section, we present our results regarding how the administrative interventions affected the
number of taxpayers. We show our main findings first for the different TREP phases, and second
for the new e-filing system.

TREP

Table 5 shows the DiD estimation results for TREP I–III. The outcome variable is the log number
of taxpayers, and we use geographical area*industry aggregated data, as explained in the previous
section. The results show that TREP I has a positive and statistically significant impact on the log
number of taxpayers. The estimate is larger when one-stop shops are established than in the earlier
years of TREP I (0.809 versus 0.473). Similarly, the estimate of TREP III is positive and significant.
TREP III has only one after-treatment period because all TREP methods, including the one-stop
shops, were introduced at the same time in July 2016. In contrast to TREP I and III, the estimates
of TREP II depend on the treatment. TREP II has a positive and statistically significant impact
on the log number of taxpayers (0.478) when we compare all post-reform years with pre-reform
years, but when we separate the after-reform variable into two, the results change. In the first years
after the introduction of TREP II, we do not find a significant impact on the log number of
taxpayers. The estimate becomes positive only after the establishment of one-stop shops (0.751).
In summary, the log number of presumptive taxpayers increases more in treatment areas than in
control areas in most of the specifications, and the largest impact, approximately 70 per cent or
more, is after the establishment of one-stop shops. The results highlight the importance of in-
person services. This is not a surprising finding in Uganda, since the National Information
Technology Survey in 2017–18 found that almost 97 per cent of individuals said they preferred to
meet government officials face to face (National Information Technology Authority Uganda
2018).

To investigate the sensitivity of the results, we run the DiD regression without cell dummies.
Columns (1), (5), and (9) of Table 5 show that the estimates are positive and significant in all
TREPs, but the size of the effect changes if cell dummies are or are not included. The estimates
are therefore slightly sensitive to additional control variables.

We visualize the development of the treatment effect in Figure 6. The left panel shows that TREP
II did not have a significant impact at the beginning, but after the introduction of one-stop shops
in July 2016 the effect becomes positive. The right panel shows that TREP III had a positive
impact from the beginning, since TREP III started with one-stop shops. However, there is already
a small positive effect before TREP III started, which might have been caused by the new e-filing
system, which was launched in July 2015. Overall, both panels support the findings from the DiD
estimation shown in Table 5. Moreover, Figure 6 verifies that in general the parallel trend
assumption holds, and we can employ the DiD approach. One caveat is that we cannot provide
visual evidence for TREP I, since it has only one pre-reform year, which is omitted if we estimate
the treatment effect by year.




                                                 16
Table 5: DiD estimation results of TREP: log number of taxpayers
                              TREP I: Kampala                            TREP II: Wakiso                TREP III: Other
                                                                                                         municipalities

                      (1)       (2)       (3)       (4)        (5)         (6)      (7)        (8)       (9)       (10)
                     After     After     After1    After2     After       After    After1     After2    After      After

TREPc*Aftert       0.881*** 0.732***    0.473** 0.809*** 0.728***       0.478**     0.193    0.751*** 0.466*** 0.779***
                   (0.174) (0.233)      (0.188) (0.288) (0.119)         (0.217)    (0.201)   (0.289) (0.077) (0.088)

Year dummies         Yes        Yes       Yes       Yes          Yes      Yes       Yes        Yes       Yes       Yes
Cell dummies         No         Yes       Yes       Yes          No       Yes       Yes        Yes       No        Yes

R-squared           0.267      0.838     0.923     0.851      0.145      0.765     0.792      0.822     0.169     0.799

Observations        3,006      3,006      377      2,714      2,823      2,823      658       2,309     3,744     3,744

                      377 taxpayers in TREP I area in             88 taxpayers in TREP II area in      1,382 taxpayers in
                                 2012/13                                     2013/14                    TREP III area in
                                                                                                            2015/16

Note: columns (1–4) present the estimates for TREP I in Kampala. Column (1) shows estimates using the after
variable, which includes all after-treatment years, not including cell dummies. Estimates in column (2) are the
same as in column (1), but with cell dummies. Column (3) presents estimates for 2013–14 to 2014–15. Column
(4) presents estimates for 2015–16 to 2017–18. Columns (5–8) show the estimates for TREP II in the Wakiso
district. Column (5) presents estimates without cell dummies using the after variable, which includes all after-
treatment years. Column (6) includes cell dummies. Columns (7) and (8) divide the after-treatment variable into
two and show the results for 2014–15 to 2015–16 and 2016–17 to 2017–18 accordingly. Columns (9) and (10)
show the estimates without and with cell dummies for TREP III in 30 municipalities using 2016–17 to 2017–18 as
after-treatment years. The number of observations is different in each column because we separately estimate
regressions for 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴1 and 𝐴𝐴𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓2 variables. Robust standard errors in parentheses. 16 *** p<0.01, ** p<0.05,
* p<0.1.
Source: authors’ calculations based on URA administrative tax returns data.




16
   In addition to robust standard errors, we estimate the DiD model using clustered standard errors at the geographical
area level as a robustness check. We cluster at the area level because it is likely that the error terms for businesses
within clusters are correlated. Clustered standard errors are on average around 60 per cent larger than robust standard
errors. However, the estimates are still significant in most of the specifications. Moreover, we use aggregated data that
eliminates business-level variation in clusters, and thus we use robust standard errors as our default specification in all
regressions in this paper.

                                                            17
Figure 6: Event study plot for TREP II and III




Note: the right panel shows the event study chart for TREP II, the left panel for TREP III. The dashed line
indicates when one-stop shops were established, and the dash-dotted line in TREP II shows when TREP started
in the Wakiso district (July 2014). The treatment effect is the coefficient of the interaction term 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔 𝑥𝑥𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑡𝑡 from
the simple DiD regression without cell dummies, but including 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔 and 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑡𝑡 dummies.
Source: authors’ calculations based on URA administrative tax returns data.

The new e-filing system

We first visualize the impact of the new e-filing system. Figure 4 shows that the number of
presumptive taxpayers increased rapidly after the introduction of the new e-filing system, whereas
there was no increase in the number of CIT payers. Figure 7 presents the development of the
estimated treatment effect using the simple DiD regression with treatment and year dummies. It
shows that the treatment effect is positive only in the second year after the introduction of the
new e-filing system. This evidence suggests that the new e-filing system increased the number of
presumptive taxpayers. Moreover, both figures show that in the pre-reform years, the treatment
and control groups have similar trends, and hence the DiD approach is suitable to measure the
size of the impact.

Table 6 presents the main findings of the DiD estimation. The results suggest that the new e-filing
system was successful: the number of presumptive taxpayers more than doubles after the
intervention when we compare all post-reform years with the pre-reform period (see columns (2–
4)). The estimates are larger when we add more control variables to the regression. In the simplest
specification, in column (1), the estimate is substantially smaller, at only 0.238. This implies that
estimates are especially sensitive to cell controls, because the estimate increases to 1.315 when we
add cell dummies (see column (2)).

We test whether the treatment effect is different for the first and second years after the reform.
We find that the effect on the log number of taxpayers is larger in the second year than in the first
year after the reform (see columns (5–10) in Table 6). Similarly for all after-years, the results are
smaller without cell dummies.




                                                              18
Figure 7: Event study plot for the new e-filing system




Note: the dashed line shows the time of the treatment (July 2015). The treatment effect is estimated from the
simple DiD regression without cell dummies. The treatment effect is the coefficient of the interaction term
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃g 𝑥𝑥𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑡𝑡 from the simple DiD regression without cell dummies but including 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃g and 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑡𝑡
dummies.
Source: authors’ calculations based on URA administrative tax return data.



Table 6: DiD estimation results of the new e-filing system: log number of taxpayers
                 (1)      (2)      (3)       (4)     (5)       (6)      (7)    (8)      (9)     (10)
                After    After    After     After   After1   After1   After1  After2  After2   After2
Presumptivec* 0.238*** 1.315*** 1.430*** 1.006*** 0.002 0.752*** 0.676*** 0.428*** 2.083*** 1.578***
Aftert         (0.078) (0.073) (0.085) (0.084) (0.093) (0.091) (0.096) (0.091) (0.109) (0.124)
TREP_Activec,t                  0.299*** -0.229***         0.362*** 0.143*           0.379*** 0.076
                                (0.052) (0.051)             (0.080) (0.082)          (0.064) (0.058)
TREP_Activec,t                            1.055***                   0.511***                 0.763***
*Presumptivec                             (0.083)                    (0.151)                  (0.137)
Presumptivec*                   0.277*** 0.138*            0.210*** 0.135*           0.299*** 0.198**
After2014/15                    (0.078) (0.077)             (0.069) (0.073)          (0.082) (0.081)
Year dummies     Yes     Yes      Yes       Yes      Yes      Yes      Yes     Yes     Yes      Yes
Cell dummies     No      Yes      Yes       Yes      No       Yes      Yes     No      Yes      Yes
R-squared       0.033   0.864    0.868     0.881    0.004    0.913    0.914   0.045   0.915    0.919
Observations    3,904   3,904    3,904     3,904    2,522    2,522    2,522   2,944   2,944    2,944
                                2,175 presumptive taxpayers in 2014-15

Note: column (1) presents basic DiD estimations for the log number of taxpayers. Column (2) shows DiD
estimates for the log number of taxpayers and includes cell dummies. Column (3) adds controls for TREP_Activec,t
and the presumptive tax rate change. Column (4) includes the TREP_Activec,t*Presumptivec interaction variable.
Columns (5–10) show results separately for the first and second years after the reform, and with and without
different control variables. After1 is the year 2015–16, and After2 is 2016–17. The number of observations differs
in each column because we separately estimate regressions for the After1 and After2 variables. Robust standard
errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Source: authors’ calculations based on URA administrative tax returns data.




                                                                     19
We control for whether TREP is active for both the treatment and control groups, and we estimate
the interaction of TREP and the treatment group. 17 In all specifications with active TREP, the
DiD estimates are positive and significant. When we estimate the interaction between active TREP
and presumptive taxpayers in column (4), the DiD estimate is smaller compared with regressions
with cell dummies or active TREP (columns (2) and (3)). The coefficient of
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 𝑥𝑥𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑐𝑐 is positive, and it is larger than the coefficient of
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇_𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑐𝑐,𝑡𝑡 . This supports findings from the analysis of TREP for presumptive taxpayers.
As a result, we argue that the new e-filing form and TREP have complementary effects, because
both increase the number of presumptive taxpayers.

The complementary effect of TREP and the new e-filing system is clearer in the second year after
the introduction of the new e-filing system. The second year is the fiscal year 2016–17, and it is
the year when the one-stop shops were established in most of the municipalities. Therefore, the
mechanism behind the complementary effect is most likely related to the one-stop shops. At a
one-stop shop, taxpayers register their businesses with different authorities one by one. They first
start their registration with the URSB. Second, they receive a TIN from URA. Finally, they can
apply for a trading licence from Kampala Capital City Authority or local government. In principle,
this system forces taxpayers to register with URA. Moreover, TREP includes client sensitization
workshops and tax education at one-stop shops, where tax officers can teach and help businesses
to file their returns using the new e-filing form. As a result, one-stop shops can have a major role
in the complementary effect.

Tax rates were reduced at the same time as the new e-filing system was introduced. We proxy
changes in taxes by adding the variable 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 𝑥𝑥𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌2014/15 , which takes value one when
the year is 2014–15, and a cell is presumptive taxpayers. In 2014–15, tax rates increased from one
to three per cent. We test whether this affected the number of taxpayers. In Table 6, the coefficient
of a proxy variable is positive, which is counter-intuitive in light of results from other studies. For
instance, Waseem (2018) found that a large increase in tax rates reduced the number of
partnerships filing returns in Pakistan; he argued that the increase led partnerships to migrate to
informality or switch to other business types. In our analysis, the positive coefficient implies that
the tax increase led to larger tax-filing. Because the coefficient is a proxy, it means that in the
opposite situation, where tax rates are reduced, the number of taxpayers should decrease—which
is opposite to the findings by Waseem (2018). Therefore, we interpret the positive coefficient as
follows: in 2014–15 the number of presumptive taxpayers slightly increased, as shown in Figure 3,
and this was a result of the positive effect of TREP I (see column (3) in Table 5). We conclude
that the tax reform did not have an impact on the number of taxpayers.

To further study the mechanism behind the increased number of presumptive taxpayers, we
examine the distribution of presumptive taxpayers and how persistently they report their incomes
to URA. Figure 8 indicates that most of the new presumptive taxpayers report the lowest possible
taxable amount of turnover, which was UGX10 million from 2015–16 onwards. Furthermore, we
investigate the persistence of presumptive taxpayers, and we find that most of the presumptive
firms report their incomes to URA only in one year and not in the next year. Overall, TREP and
the new e-filing system successfully enhance the formalization of businesses, but they do not
manage to keep small businesses on board. The reason for this might be the low enforcement



17
  We control for active TREP for the control group because when we analyse the effect of TREP for small CIT
payers using a similar model to Equation [1], we find mixed results. The results for TREP I and II are small, positive,
and significant, but for TREP III the estimate is not significant.

                                                         20
capacity of tax officials. In general, enforcement operations in low-income countries with limited
administrative capacity focus on large firms, from which most of the tax revenues naturally arise.

We test the robustness of the estimation results using different groups of CIT payers as a control
group. The tested groups are CIT payers with turnovers of (1) UGX0–150 million, (2) over UGX0
but less than UGX150 million, and (3) UGX0–400 million. The results do not change significantly
in any different specification of the control group, and the estimate is around one for all groups
when we compare all pre-reform years with post-reform years. Figure 3 supports this finding,
because it shows that the growth of CIT payers whose turnover is between UGX0 and UGX400
million was constant from 2011–12 to 2016–17.
Figure 8: Distribution of presumptive taxpayers, 2012–13 to 2017–18




Note: vertical lines mark the thresholds of tax brackets in fiscal years. The size of the bin is UGX100,000. All
monetary values are in UGX.
Source: authors’ calculations based on URA administrative tax returns data.

5.2    Tax revenues

The objectives of TREP and the new e-filing system were mainly to formalize businesses and make
taxpaying easier, thus reducing compliance costs. As a result of increased formalization, the tax
revenues collected from small businesses might have increased. In this section, we evaluate three
possible channels through which presumptive tax revenues might have increased. First, higher
presumptive tax rates might increase tax revenues. Second, the presumptive tax base might have
been expanded. Third, more small businesses might have become taxpayers.

To evaluate these possible channels, we first examine the impacts of TREP and the new e-filing
system on payable taxes, using a similar DiD estimation as in section 5.1, and we provide

                                                         21
descriptive evidence of how presumptive tax revenues developed between 2012–13 and 2017–18.
Second, we discuss the potential increase in the presumptive tax base.

We evaluate the impact of TREP and the new e-filing system on payable taxes as these are observed
in the returns data. Table 7 shows that the results of the different TREPs are significant in most
of the specifications. Columns (1), (5), and (9) present the estimates when we do not add cell
dummies. These estimates are significant and vary from 0.5 for TREP III to 1.26 for TREP I.
When we include cell dummies, the results change: the estimate of TREP I drops to 1.02, TREP
II becomes insignificant, and TREP III rises to 0.784. Further, we separately investigate the effects
of TREP before and after the establishment of the one-stop shops by dividing the post-TREP
period into two for TREP I and II. In both TREPs, the estimates are larger for the years when
one-stop shops were in place. In summary, our results show that TREP increased presumptive tax
revenues by approximately 70 per cent or more in most of the specifications.
Table 7: DiD estimation results for TREP: log payable taxes

                           TREP I: Kampala                       TREP II: Wakiso              TREP III: Other
                                                                                               municipalities
                    (1)      (2)    (3)     (4)      (5)    (6)    (7)     (8)                  (9)       (10)
                   After    After  After1 After2    After  After  After1  After2               After     After
TREPc*Aftert     1.255*** 1.021*** 0.210 1.216*** 0.977*** 0.501 0.250 0.746*                0.502*** 0.784***
                 (0.233) (0.358) (0.377) (0.388) (0.193) (0.334) (0.498) (0.402)             (0.092)    (0.100)
Year dummies       Yes      Yes     Yes    Yes      Yes     Yes    Yes     Yes                 Yes        Yes
Cell dummies        No      Yes     Yes    Yes       No     Yes    Yes     Yes                  No        Yes
R-squared         0.302    0.849   0.899  0.861    0.143   0.787 0.860    0.843               0.180      0.815
Observations      2,919    2,919    290   2,698    2,703   2,703   538    2,257               3,607      3,607

Note: columns (1–4) present the estimates for TREP I in Kampala. Column (1) shows estimates using the after
variable, which includes all after-treatment years, not including cell dummies. In column (2), we add cell
dummies. Column (3) presents estimates for 2013–14 to 2014–15. Column (4) presents estimates for 2015–16 to
2017–18. Columns (5–8) show the estimates for TREP II in the Wakiso district. Column (5) presents estimates
without cell dummies using the after variable, which includes all after-treatment years. Column (6) shows the
same estimates as column (5), but with cell dummies. Columns (7) and (8) divide the after-treatment variable into
two and show the results for 2014–15 to 2015–16 and 2016–17 to 2017–18 accordingly. Columns (9) and (10)
show the estimates for TREP III in 30 municipalities, using 2016–17 to 2017–18 as the after-treatment years
without and with cell dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Source: authors’ calculations based on URA administrative tax returns data.

In Table 8, we show similar DiD estimates of payable taxes for the new e-filing system. In columns
(1–4) we estimate the overall effect for post-reform years, and in columns (5–10) we look
separately at impacts for the first and second years after the reform. These results are sensitive to
different control variables. When we do not include cell dummies, the estimate is negative and/or
insignificant. After we include cell dummies, control for active TREP, or include the interaction
between active TREP and the treatment group, the estimates change to positive (around 1.0). A
similar pattern is seen when we look separately at the first and second years after the reform. The
largest positive effect—1.65–2.1, depending on the specification—is for the second year after the
introduction of the new e-filing system. Overall, the new e-filing system with TREP increased
presumptive tax revenues (see columns (4), (7), and (10)).




                                                       22
Table 8: DiD estimation results for the new e-filing system: log payable taxes

                 (1)     (2)      (3)      (4)     (5)      (6)      (7)    (8)      (9)     (10)
                After   After    After    After  After1   After1   After1  After2  After2   After2
Presumptivec* -0.120 1.138*** 1.407*** 0.996*** -0.359** 0.736*** 0.587*** 0.086 2.087*** 1.649***
Aftert         (0.131) (0.125) (0.148) (0.155) (0.162) (0.173) (0.183) (0.152) (0.199) (0.253)
TREP_Activec,t                 0.368*** -0.209           0.419** 0.041            0.426*** 0.147
                               (0.088) (0.130)           (0.164) (0.215)          (0.143) (0.200)
TREP_Activec,t                          1.038***                  0.934***                 0.668**
*Presumptivec                           (0.146)                   (0.268)                  (0.265)
Presumptivec*                  0.629*** 0.481***         0.624*** 0.471***        0.578*** 0.484***
After2014/15                   (0.155) (0.155)           (0.147) (0.157)          (0.170) (0.171)
Year dummies     Yes     Yes     Yes      Yes     Yes      Yes      Yes     Yes     Yes      Yes
Cell dummies     No      Yes     Yes      Yes      No      Yes      Yes     No      Yes      Yes
R-squared       0.245   0.861   0.864    0.869   0.269    0.882    0.883   0.220   0.892    0.893
Observations    3,182   3,182   3,182    3,182   1,933    1,933    1,933   2,342   2,342    2,342

Note: column (1) presents simple DiD estimations without cell control variables. Column (2) shows DiD estimates
including cell dummies. Column (3) adds controls for TREP_Activec,t and the presumptive tax rate change, and
column (4) for TREP_Activec,t*Presumptivec. Columns (5–10) show results separately for the first and second
years after the reform, without and with different control variables. After1 is the year 2015–16, and After2 is 2016–
17. The number of observations is different in each column because we separately estimate the regressions for
the After1 and After2 variables. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Source: authors’ calculations based on URA administrative tax returns data.

Figure 9: Tax revenues calculated from presumptive tax returns data, 2011–12 to 2017–18




Note: tax revenues are calculated from variable payable taxes. All monetary values are in UGX.
Source: authors’ calculations based on URA administrative tax returns data.

Figure 9 visualizes the trend in tax revenues calculated from presumptive tax returns data from
2011–12 to 2017–18. This tax revenue represents the revenues if all payable taxes in tax returns
had been paid. The tax revenue increases substantially after 2015–16. Figure 9 is remarkably similar
to the left side of Figure 3, which shows the trend in the number of presumptive taxpayers.
Therefore, we argue that the increase in presumptive tax revenues is mainly due to the increased
number of taxpayers, rather than to any changes in presumptive tax policy (in the next section we
investigate this in more detail). We showed in section 5.1 that both TREP and the new e-filing
system increased the number of presumptive taxpayers. Moreover, the estimation results in this
section suggest that the amount of payable taxes increased in active TREP areas and after the
introduction of the new e-filing form.

                                                         23
One possible reason for the increased tax revenues is the wider presumptive tax base. In 2015–16,
the upper threshold for presumptive tax increased from UGX50 million to UGX150 million.
Figure 8 presents the distribution of presumptive taxpayers from 2012–13 to 2017–18, and it
shows that the number of presumptive taxpayers between UGX50 million and UGX150 million
is very low in all years. Therefore it is reasonable to assume that the tax revenues shown in Figure
9 did not increase due to the change in the upper threshold alone.

Robustness checks

In the analysis of tax revenues, we need to take into account the presumptive tax rate reforms in
2014–15 and 2015–16. In 2014–15, the presumptive tax rate changed from one to three per cent;
in 2015–16 the tax rates were reduced on average to 1.5 per cent. Therefore, as a robustness check,
we estimate the effect on payable taxes, which we calculate from estimated sales using a
hypothetical marginal tax rate (MTR) of 1.7 per cent for every year. This hypothetical MTR is an
average presumptive tax rate for all years. We also use UGX10 million as a zero-tax threshold.

In Tables 9 and 10, we show the results of TREP and the new e-filing form using the hypothetical
MTR for all years. The results are similar to those found in the previous section, when we did not
control for the change in marginal tax rates (see Tables 7 and 8). Therefore, we can conclude that
the presumptive tax rate changes had no effect on presumptive tax revenues.
Table 9: Robustness check for TREP: DiD estimation results of log payable taxes when the marginal tax rate is
the same for all years

                           TREP I: Kampala                       TREP II: Wakiso              TREP III: Other
                                                                                               municipalities
                    (1)     (2)    (3)     (4)     (5)     (6)     (7)     (8)                  (9)       (10)
                   After   After  After1  After2  After   After   After1  After2               After     After
TREPc*Aftert     0.924*** 0.763* 0.154 0.834** 0.924*** 0.315 0.070       0.591              0.605*** 0.982***
                 (0.234) (0.398) (0.486) (0.414) (0.195) (0.389) (0.491) (0.466)             (0.093)    (0.116)
Year dummies       Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes                 Yes        Yes
Cell dummies        No     Yes     Yes     Yes     No      Yes     Yes     Yes                  No        Yes
R-squared         0.193   0.824   0.873   0.837   0.109   0.788 0.861     0.839               0.142      0.815
Observations      2,730   2,730    265    2,525   2,512   2,512    370    2,213               3,393      3,393

Note: columns (1–4) present the estimates for TREP I in Kampala. Column (1) shows estimates using the after
variable, which includes all after-treatment years, not including cell dummies. In column (2), we add cell
dummies. Column (3) presents estimates for 2013–14 to 2014–15. Column (4) presents estimates for 2015–16 to
2017–18. Columns (5–8) show the estimates for TREP II in the Wakiso district. Column (5) presents estimates
without cell dummies using the after variable, which includes all after-treatment years. Column (6) shows the
same estimates as column (5), but with cell dummies. Columns (7) and (8) divide the after-treatment variable into
two and show results for the years 2014–15 to 2015–16 and 2016–17 to 2017–18 accordingly. Columns (9) and
(10) show the estimates for TREP III in 30 municipalities using 2016–17 to 2017–18 as the after-treatment years,
without and with cell dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Source: authors’ calculations based on URA administrative tax returns data.




                                                       24
Table 10: Robustness check for the new e-filing system: DiD estimation results of log payable taxes when the
marginal tax rate is the same for all years

                   (1)     (2)      (3)      (4)     (5)           (6)          (7)       (8)      (9)     (10)
                  After   After    After    After   After1        After1       After1    After2  After2   After2
Presumptivec*     0.060 1.188*** 1.176*** 0.783*** -0.250        0.268*       0.244*    0.299** 2.088*** 1.812***
Aftert           (0.130) (0.132) (0.131) (0.135) (0.160)         (0.143)      (0.145)   (0.151) (0.180) (0.232)
TREP_Activec,t                   0.350*** -0.246*                 0.237        0.127            0.386*** 0.205
                                 (0.099) (0.135)                 (0.161)      (0.208)           (0.147) (0.199)
TREP_Activec,t                            1.133***                             0.272                      0.452*
*Presumptivec                             (0.153)                             (0.264)                    (0.269)
Year dummies       Yes     Yes     Yes      Yes      Yes           Yes          Yes       Yes     Yes      Yes
Cell dummies       No      Yes     Yes      Yes      No            Yes          Yes       No      Yes      Yes
R-squared         0.155   0.833   0.835    0.842   0.185          0.861        0.861     0.132   0.875    0.875
Observations      3,008   3,008   3,008    3,008   1,775          1,775        1,775     2,300   2,300    2,300

Note: column (1) presents simple DiD estimations without cell control variables. Column (2) shows DiD estimates
including cell dummies. Column (3) adds controls for TREP_Activec,t, and column (4) for
TREP_Activec,t*Presumptivec. Columns (5–10) show results separately for the first and second years after the
reform, without and with different control variables. After1 is the year 2015–16, and After2 is 2016–17. The
number of observations is different in each column because we separately estimate the regressions for the After1
and After2 variables. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Source: authors’ calculations based on URA administrative tax returns data.

Figure 10: Tax revenues calculated from presumptive tax returns data from 2012–13 to 2017–18, for taxpayers
with turnover UGX0–50 million only




Note: tax revenues are calculated from variable payable taxes. All monetary values are in UGX.
Source: authors’ calculations based on URA administrative tax return data.

Altogether, the increase in presumptive tax revenues shown in Figure 9 is thus mainly due to the
increase in the number of taxpayers. The DiD estimation results in Tables 7 and 8 show that tax
revenues from presumptive taxpayers whose turnover is UGX0–50 million increased because of
TREP and the new e-filing system. Furthermore, the results of the robustness checks in Tables 9
and 10 support this argument, because they show that presumptive tax reforms do not affect the
results. On the other hand, widening the presumptive tax base in 2015–16 naturally increased
presumptive tax revenues; but as shown in Figure 8, only a few new presumptive taxpayers had a
turnover between UGX50 million and UGX150 million. Moreover, Figure 10 supports this
argument, because it shows that the increase in tax revenues is mainly due to presumptive taxpayers


                                                      25
whose turnover is between UGX0 and UGX50 million, and therefore the main result is that the
increase in tax revenues is due to the new e-filing system and TREP.

5.3       A cost-benefit analysis of TREP

In this section, we analyse the costs and benefits of TREP. First, we compare presumptive tax
revenues in active TREP and control areas for the years before and after TREP. This DiD
represents the additional presumptive tax revenue assessed because of TREP. We use this
simplified approach to calculate the additional revenue because we use a similar approach to
calculate the cost of TREP. Second, we calculate the average costs of TREP from URA’s budget
for TREP. Finally, we discuss whether TREP has benefits not only for the revenue authority but
also for businesses, and how cost-effective it is.

In Table 11, we show revenue calculations for TREP I and III. We only calculate tax revenues for
TREP I and III because TREP II did not have a significant impact on payable taxes (see Table 9).
The average additional tax revenue from TREP I is around UGX11.4 million and from TREP III
almost UGX3 million per year and cell. 18 However, the additional tax revenue from TREP III may
be upwardly biased, because Figure 6 shows an increasing trend in pre-reform years. To sum up,
the average additional tax revenues from TREP per year and cell are around UGX14 million.
Table 11: Average tax revenues per cell in TREP I and III areas

                                                                            TREP I          Control        Difference
Before (2012-13 to 2014-15)                                               7,282,978         503,738        6,779,240
After (2015-16 to 2017-18)                                                42,259,840       1,373,424       40,886,416
Difference                                                                34,976,862        869,686        34,107,176
Yearly average                                                                                             11,369,059
                                                                            TREP III        Control        Difference
Before (2012-13 to 2015-16)                                                1,477,894        392,476        1,085,418
After (2016-17 to 2017-18)                                                 8,146,134       1,342,768       6,803,366
Difference                                                                 6,668,240        950,292        5,717,498
Yearly average                                                                                             2,858,974
Average additional revenue from TREP I and III per year                                                    14,228,033

Note: average revenues are calculated for presumptive taxpayers whose turnover is under UGX50 million.
Taxpayers who are in TREP I and III areas only are included in this analysis because those areas have a
significant increase in tax revenues based on the DiD estimations (see Table 9). All monetary values are in UGX.
Source: authors’ calculations based on URA administrative tax return data.

Next, we present average expenditures on all TREP areas in Table 12. The calculated average yearly
expenditure on TREP per cell is around UGX1.5 million, which is only 11 per cent of the average
additional tax revenue per year and cell.




18
     A cell here refers to the same cell variable (=geographical area*industry) as we have used in previous sections.

                                                            26
Table 12: Average expenditure per cell on all TREP areas

                                                         Average expenditure on TREP per cell
2015-16                                                              1,861,751
2016-17                                                              1,143,223
2017-18                                                              1,618,082
Yearly average                                                       1,541,019

Note: average expenditure is calculated by dividing the TREP budget by the number of cells per year. 19 All
monetary values are in nominal UGX.
Source: authors’ calculations based on internal URA documents.

Our cost-benefit analysis presents only simplified back-of-the-envelope calculations. We do not
use a similar DiD approach as in previous sections, because estimates from regressions would be
harder to compare with the simple costs of TREP. Moreover, we do not have information
regarding how costs are divided and utilized in different TREP municipalities, since we know only
the total expenditure on TREP. The simplified analysis illustrates that the average additional
revenue gained from TREP is much higher than the sum that URA has spent on TREP yearly.
Against this background, TREP has been a successful intervention for increasing presumptive tax
revenues. However, when we look at the bigger picture, presumptive tax revenues are a minor part
of all tax revenues in Uganda. Hence, the main objective of TREP was to formalize businesses—
not to increase revenue collection—which URA has done cost-effectively.

One caveat on our analysis is that we only have information about tax returns and not tax
payments. The problem is that taxes paid are usually lower than tax assessed in tax returns. For
example, in URA’s official statistics, the presumptive tax collection in 2016–17 was UGX4.46
billion, but when we calculate payable presumptive taxes using the returns data, the amount is
UGX5.54 billion (URA 2017). The difference between payments and returns is around 20 per cent.
If we take this into consideration and lower the presumptive tax revenue by 20 per cent, the average
additional tax revenue from TREP I and III is around UGX11 million. This value is still
substantially larger than the costs of TREP.

Although TREP is mainly targeted at small businesses, there might have been spillover effects on
medium-sized and large corporations, and on individual businesses such as the self-employed.
Many methods used in TREP are public, such as radio advertisements and newspaper messages,
and they can reach many citizens. Therefore, the expenditure on TREP is not only used to
formalize small businesses. This means that TREP may have been even more cost-effective than
we have calculated.

Besides its benefits for the revenue authority, TREP also has benefits for taxpayers. For taxpayers,
TREP and especially the one-stop shops have enabled easier access to the services of government
agencies, which saves taxpayers time and money. The 2017–18 National Information Technology
Survey found that 96.6 per cent of individuals preferred to meet government officials in person
(National Information Technology Authority Uganda 2018). Thus one-stop shops provide the
face-to-face services that citizens demand. In addition, taxpayers benefit from formalization
because when they become formalized, they gain access to financial and business services, and
public goods such as social security. Moreover, TREP enhances the taxpaying culture in Uganda
by informing people about taxes, which might benefit the country in a broader way.


19
  The TREP budget is in nominal values, and we do not take inflation into account. Inflation was 4.78 per cent per
year on average from 2014 to 2018 (World Bank 2020a). If we were to make inflation adjustments, the average yearly
expenditure would become lower.

                                                       27
Another caveat on our cost-benefit calculation for TREP is that we only take account of URA’s
expenditure on TREP as a cost. However, other costs can arise for taxpayers when they become
formalized, because they need to pay taxes, which lowers their disposable income. This cost-
benefit analysis does not value the lost consumption of taxpayers, and therefore the costs of TREP
may be larger than we have calculated.

This cost-benefit analysis could be used as a starting point for calculating the optimal enforcement
elasticity of tax revenue. Keen and Slemrod (2017) introduced the concept of the enforcement
elasticity of tax revenue, and built a theoretical and empirical framework for analysing the
optimality of administrative interventions. The cost-benefit ratio—around 1:10, in our
estimation—is the first part of the enforcement elasticity; the second part would be the compliance
costs estimates. Most likely TREP has reduced taxpayers’ compliance costs, but the estimation of
compliance costs is beyond the scope of this paper.


6     Conclusion


Small businesses are a large and important part of the economy in developing countries, but they
mostly operate informally. The formalization of small businesses is a potential source for revenue
authorities to increase domestic revenue mobilization. However, reaching small businesses,
convincing them to become formalized, an enforcing that formalization is a difficult task.

In this study we evaluate the impact of two administrative interventions, TREP and a new
electronic filing system for presumptive tax, on the number of small businesses and on revenue
collection, using the DiD method. We use Ugandan administrative tax data covering both
presumptive and CIT payers from 2012–13 to 2017–18. Furthermore, we conduct a cost-benefit
analysis of TREP using simplified back-of-the-envelope calculations and URA’s TREP budget.

We find that both TREP and the new e-filing system substantially increased the number of
presumptive taxpayers, and almost doubled the presumptive tax revenue collection. Our results
suggest that these reforms have had complementary effects, because at one-stop shops taxpayers
cannot avoid registering with URA, and tax officers can help taxpayers to file their returns using
the new e-filing system. Overall, our findings highlight that both TREP and the new e-filing system
have enhanced the formalization and tax compliance of small firms in Uganda.

The results of the cost-benefit analysis show that the additional average tax revenue from TREP
was almost ten times higher than URA’s yearly TREP budget. In our analysis, we showed that this
result holds even if we use payment data instead of returns data. Moreover, in light of TREP’s
other benefits, such as its potential spillover effects on medium-sized businesses and the
improvement of taxpayers’ access to services, we conclude that TREP is a highly cost-effective
initiative.

The adoption of advanced technologies in tax administration is a promising development for low-
income countries. Technologies can improve weak administrative capacity by reducing costs and
increasing efficiency. In addition, better services and more inclusive methods to enhance
formalization can complement technological improvements. We have shown that a large-scale
taxpayer registration programme targeted at small businesses, complemented by simplified tax-
filing, may be a cost-effective intervention to increase tax compliance and revenue collection in
Uganda. Further studies of similar interventions in developing countries using administrative data
would be needed to generalize the results of this paper to other countries.


                                                28
References


Alm, J. (2019). ‘What Motivates Tax Compliance?’ Journal of Economic Surveys, 33(2): 353–88.
    https://doi.org/10.1111/joes.12272
Benhassine, N., D. McKenzie, V. Pouliquen, and M. Santini (2018). ‘Does Inducing Informal Firms to
    Formalize Make Sense? Experimental Evidence from Benin’. Journal of Public Economics, 157: 1–14.
    https://doi.org/10.1016/j.jpubeco.2017.11.004
Besley, T., and T. Persson (2014). ‘Why Do Developing Countries Tax So Little?’ Journal of Economic
     Perspectives, 28(4): 99–120. https://doi.org/10.1257/jep.28.4.99
Best, M.C., A. Brockmeyer, H.J. Kleven, J. Spinnewijn, and M. Waseem (2015). ‘Production Versus
     Revenue Efficiency with Limited Tax Capacity: Theory and Evidence from Pakistan’. Journal of Political
     Economy, 123(6): 1311–55. https://doi.org/10.1086/683849
Brockmeyer A., S. Smith, M. Hernandez, and S. Kettle (2019). ‘Casting a Wider Tax Net: Experimental
    Evidence from Costa Rica’. American Economic Journal: Economic Policy, 11(3): 55–87.
    https://doi.org/10.1257/pol.20160589
Bruhn, M., and D. McKenzie (2014). ‘Entry Regulation and the Formalization of Microenterprises in
    Developing      Countries’.   World     Bank      Research  Observer,     29(2):    186–201.
    https://doi.org/10.1093/wbro/lku002
Carillo, P., D. Pomeranz, and M. Singhal (2017). ‘Dodging the Taxman: Firm Misreporting and Limits to
      Tax Enforcement’. American Economic Journal: Applied Economics, 9(2): 144–64.
      https://doi.org/10.1257/app.20140495
Floridi, A., B.A. Demena, and N. Wagner (2020). ‘Shedding Light on the Shadows of Informality: A Meta-
     Analysis of Formalization Interventions Targeted at Informal Firms’. Labour Economics, 67: 101925.
     https://doi.org/10.1016/j.labeco.2020.101925
International Telecommunication Union (2020). ‘ICT-Eye ICT Data Portal’. Available at:
     www.itu.int/net4/ITU-D/icteye#/ (accessed 1 December 2020).
Keen, M., and J. Slemrod (2017). ‘Optimal Tax Administration’. Journal of Public Economics, 152: 133–42.
    https://doi.org/10.1016/j.jpubeco.2017.04.006
Lediga, C., N. Riedel, and K. Strohmaier (2020). ‘What You Do (and What You Don’t) Get When
     Expanding the Net: Evidence from Forced Taxpayer Registrations in South Africa’. SSRN.
     http://dx.doi.org/10.2139/ssrn.3616808
Mascagni, G. (2018). ‘From the Lab to the Field: A Review of Tax Experiments’. Journal of Economic Surveys,
    32(2): 273–301. https://doi.org/10.1111/joes.12201
Mascagni, G., F. Santoro, and D. Mukama (2019). ‘Teach to Comply? Evidence from a Taxpayer Education
    Programme in Rwanda’. ICTD Working Paper 91. Brighton: Institute of Development Studies.
    Available at: www.ictd.ac/publication/teach-comply-evidence-taxpayer-education-programme-
    rwanda/ (accessed 13 October 2020).
National Information Technology Authority Uganda (2018). ‘National Information Technology Survey:
     2017–18 Report’. Kampala: National Information Technology Authority Uganda. Available at:
     www.nita.go.ug/publication/national-it-survey-2018-final-report (accessed 21 July 2020).
Okunogbe, O., and V. Pouliquen (2018). ‘Technology, Taxation, and Corruption: Evidence from the
    Introduction of Electronic Tax Filing’. Policy Research Working Paper 8452. Washington, DC: World
    Bank. Available at: https://openknowledge.worldbank.org/handle/10986/29862 (accessed 23
    January 2019).
Uganda Bureau of Statistics (2016). ‘Urban Labour Force Survey 2015: Fact Sheet’. Kampala: Uganda
    Bureau          of           Statistics.      Available         at:         www.ubos.org/wp-
    content/uploads/publications/03_2018ULFS_2015_Fact_Sheet_(1).pdf (accessed 8 July 2020).


                                                    29
Uganda Bureau of Statistics (2019). ‘2019 Statistical Abstract’. Kampala: Uganda Bureau of Statistics.
    Available at: www.ubos.org/wp-content/uploads/publications/01_20202019_Statistical_Abstract_-
    Final.pdf (accessed 23 April 2020).
URSB (2017). ‘Annual Report 2016–17’. Kampala: URSB. Available at: https://ursb.go.ug/wp-
   content/uploads/2018/08/Annual-Report-2016-2017-1.pdf (accessed 23 October 2018).
URA (2015). Taxation Handbook: A Guide to Taxation in Uganda. Kampala: URA. Available at:
   www.ura.go.ug/Resources/webuploads/INLB/TAXATION%20HANDBOOK%20.pdf (accessed
   11 August 2017).
URA      (2017). ‘June Revenue Performance Report’. Kampala: URA. Available at:
      www.ura.go.ug/Resources/webuploads/GNRART/Annual%20Revenue%20Report%20%202016_
      17.pdf (accessed 8 August 2018).
UNESCO Institute for Statistics (2020). ‘UIS Database’. Available at: http://data.uis.unesco.org/ (accessed
   21 July 2020).
UNU-WIDER (2020). ‘UNU-WIDER Government Revenue Dataset’. Available at:
   https://www.wider.unu.edu/project/government-revenue-dataset (accessed 24 November 2020).
Waseem, M. (2018). ‘Taxes, Informality and Income Shifting: Evidence from a Recent Pakistani Tax
    Reform’. Journal of Public Economics, 157: 41–77. https://doi.org/10.1016/j.jpubeco.2017.11.003
Waseem, M. (2020). ‘The Role of Withholding in the Self-Enforcement of a Value-Added Tax: Evidence
    from Pakistan’. Review of Economics and Statistics. https://doi.org/10.1162/rest_a_00959
World Bank (2017). ‘From Regulations to Enablers: Role of City Governments in Economic Development
    of     Greater     Kampala’.      Washington,       DC:      World     Bank.    Available   at:
    https://openknowledge.worldbank.org/handle/10986/28459 (accessed 21 July 2020).
World Bank (2018). ‘GDP Per Capita, PPP (Current International $): Uganda’. Available at:
    https://data.worldbank.org/indicator/NY.GDP.PCAP.PP.CD?locations=UG (accessed 26 October
    2018).
World Bank (2020a). ‘Inflation, Consumer Prices (Annual %): Uganda’. Available at:
    https://data.worldbank.org/indicator/FP.CPI.TOTL.ZG?end=2019&locations=UG (accessed 8
    January 2021).
World      Bank      (2020b).    ‘Kampala      Informal     Sector     Survey’.    Available  at:
    https://datacatalog.worldbank.org/dataset/kampala-informal-sector-survey (accessed 8 January
    2021).
Yilmaz, F., and J. Coolidge (2013). ‘Can E-Filing Reduce Tax Compliance Costs in Developing Countries?’
    Policy Research Working Paper 6647. Washington, DC: World Bank. Available at:
    https://openknowledge.worldbank.org/handle/10986/16861 (accessed 15 November 2018).


Legislation and regulations


The Income Tax Act, of 1 July 1997, Cap 340 of the Laws of Uganda 2000.
The Income Tax (Amendment) Act 2003, of 1 July, The Uganda Gazette, Volume XCVI, No. 3, of 17
     January 2004.
The Income Tax (Amendment) Act 2014, of 1 July, The Uganda Gazette, Volume CVII, No. 61, of 24
     October 2014.
The Income Tax (Amendment) Act 2015, of 1 July, The Uganda Gazette, Volume CVIII, No. 38, of 10
     July 2015.
The Income Tax (Amendment) Act 2016, of 1 July, The Uganda Gazette, Volume CIX, No. 88, of 30
    December 2016.

                                                    30
Value Added Tax Act, of 1 July 1996, Cap 349 of the Laws of Uganda 2000.
Value Added Tax (Amendment) Act 2015, of 1 July, The Uganda Gazette, Volume CVIII, No. 30, of 5
     June 2015.
Value Added Tax (Rate of Tax) Order 2006, of 1 July 2005, Under Section 78 of the Value Added Tax Act,
     Statutory Instruments 2006 No. 29, The Uganda Gazette, Volume XCVIX, No. 39, of 30 June 2006.




                                                 31
Appendix A

Table A1: Tax rates for small business taxpayers in Uganda, 2002–03 to 2017–18

Fiscal year                  Effective from                 Turnover in millions              Tax rate
2015-16 to 2017-18           1 July 2015                           0-10                         0%
                                                                  10-50                          ***
                                                                  50-75                   937,500 or 1.5 %
                                                                 75-100                  1,312,500 or 1.5%
                                                                 100-125                 1,687,500 or 1.5%
                                                                 125-150                 2,062,500 or 1.5%
2014-15                      1 July 2014                           0-10                         0%
                                                                  10-20                    450,000 or 3%
                                                                  20-30                    750,000 or 3%
                                                                  30-40                   1,050,000 or 3%
                                                                  40-50                   1,350,000 or 3%
2003-04 to 2013-14           1 July 2002                            0-5                         0%
                                                                   5-20                       100,000
                                                                  20-30                    250,000 or 1%
                                                                  30-40                    350,000 or 1%
                                                                  40-50                    450,000 or 1%

Note: tax rates for turnover between UGX10 million and UGX50 million are lump sum amounts, presented in
Table A2. These tax rates depend on business sector and location of business. The effective tax rates are either
a lump sum or percentage amount, whichever is lower. The upper value of a tax bracket is included in that tax
bracket; when it is exceeded, the taxpayer is switched to the next tax bracket. All monetary values are in UGX.
Source: authors’ interpretation of the Income Tax Act (of 1 July 1997, including Income Tax (Amendment) Acts
2003, 2014, 2015, and 2016, all of 1 July).




                                                       32
Table A2: Tax rates as lump sum amounts for small business taxpayers for different turnover brackets, locations,
and business sectors, 2015–16 to 2017–18

  Location              Business sector                    Turnover             Turnover            Turnover
                                                         10-20 million        20-35 million        35-50 million
  Kampala city          General trade                      250,000              400,000              500,000
  and divisions of      Carpentry/metal workshops          250,000              400,000              500,000
  Kampala
                        Garages                            300,000              450,000              550,000
                        Hair and beauty salons             300,000              400,000              550,000
                        Restaurants or bars                300,000              450,000              550,000
                        Drug shops (=pharmacies)           250,000              350,000              500,000
                        Others                             200,000              300,000              450,000
  Municipalities        General trade                      150,000              300,000              400,000
                        Carpentry/metal workshops          150,000              300,000              400,000
                        Garages                            200,000              350,000              450,000
                        Hair and beauty salons             200,000              350,000              450,000
                        Restaurants or bars                200,000              350,000              450,000
                        Drug shops                         150,000              300,000              400,000
                        Others                             150,000              350,000              400,000
  Towns and             General trade                      100,000              200,000              300,000
  trading               Carpentry/metal workshops          100,000              200,000              300,000
  centres
                        Garages                            100,000              250,000              350,000
                        Hair and beauty salons             100,000              250,000              350,000
                        Restaurants or bars                100,000              250,000              350,000
                        Drug shops                         100,000              200,000              300,000
                        Others                             100,000              250,000              300,000

Note: the upper value of a tax brackets is included in that tax bracket; when it is exceeded, the taxpayer is
switched to the next tax bracket. All monetary values are in UGX.
Source: authors’ interpretation of the Income Tax (Amendment) Acts (2015, 2016, both of 1 July).




                                                        33
Appendix B: TREP working methods


TREP includes several different working methods to sensitize taxpayers. Here we briefly explain
what these different methods are.
One-stop shops are service centres that include representatives from URA, URSB, Kampala
Capital City Authority, and local governments. In general, one-stop shops are located in local
government offices in municipalities and districts, but there are also one-stop shops at URA’s
offices. All one-stop shops have at least one officer from each agency (i.e. URA, URSB, and local
government), but there are more officers in busy locations. The one-stop shop is equipped with
service desks, computers, and barcode readers that are used to complete registrations. The one-
stop shops usually serve local businesses, but a client from any part of the country can visit and
register in any one-stop shop.
Stakeholder engagements are meetings between TREP officers and local councils and leaders. In
the engagement sessions, TREP officers explain the kinds of activities they perform within their
localities. The objective of the meetings is to win the support of local councils and leaders.
Quarterly messages to staff are messages about URA’s recent developments, which URA
headquarters sends quarterly to other URA offices.
Client sensitization workshops are meetings where URA tax officers educate businesses and
individuals on how to file and pay taxes.
Public notices are news items giving information and guidelines, which URA’s media teams
produce for publication in newspapers and social media. These give more detailed information
about taxes and other obligations.
Newspaper strips are shorter messages or news items with a special focus on one topic. They are
published several times or regularly in newspapers. They are like advertisements, but informative.
Barazas are organized public meetings (larger events) where URA officers educate taxpayers about
certain taxes, and afterwards participants can ask questions related to the topic. These events
provide a platform for taxpayers to air their concerns, ask questions, etc.
Radio advertisements are short advertisements on local radio stations in local languages.
Public van communications are advertisements on public vans.
Radio talk shows are talk shows about registration and taxes, on both local and national stations.
TV scripts are like advertisements but longer, and are shown on local and national TV channels.
Press briefings are events that URA organizes to inform the press.
Door-to-door registration of businesses involves TREP officers going door to door and issuing
tax assessments. At this point, business owners that refuse to cooperate are registered for taxes by
force. However, URA has no legal authority to close a business without a TIN, since URA can
only close a business after issuing tax assessments and if the taxpayer has failed to pay. In contrast,
local governments can close a business if it does not have a business licence. URA and local
government officers go from door to door together, and any business found without a business
licence is closed and directed to go to a one-stop shop to get a licence. Before the licence is given,
the business is first required to register for a TIN and pay the taxes assessed.




                                                  34


File and source

File

as-of-april-2021-the-agency-resumed-collecting-tax-return-data-from-the-internal-revenue-service-irs-an-oig-recommendati.pdf

Size

1,466,582 bytes

SHA-256

c0770e11bffec8a6d48dfca988de02fa46c2fe88dd52dd21e7d1e7a76a6373b8

Our copy

as-of-april-2021-the-agency-resumed-collecting-tax-return-data-from-the-internal-revenue-service-irs-an-oig-recommendati.pdf

Original

www.econstor.eu