QuBE – ATO Report

Cash and the Hidden Economy - Fighting Tax

Evasion in Small Business Transactions:

An Experimental Approach

Prepared by:

Uwe Dulleck, Jonas Fooken, Naomi Moy, and Benno Torgler

1 | P a g e

2 | P a g e

Executive Summary Increasing the tax compliance of small business remains an ongoing challenge for tax

authorities. From a compliance point of view, cash transactions are particularly problematic

when services are paid for on the spot, as such exchanges are difficult to audit as cash

transaction are easier for businesses not to report. In order to reduce the number of cash

transactions, increase the paper trail, and consequently increase compliance, several

behavioural economic policies can be explored. However, before scaling it up in the field, the

mechanisms through which the policies work can be tested in the laboratory so as to guide

implementation. Thus, laboratory experiments can provide some first pre-tests of the

mechanism of compliance. This report summarises laboratory-based tests of several

compliance measures.

The measures to increase compliance described in this report are the following: (1) the

Australian Tax Office (ATO) offering assistance to businesses, (2) conducting a clamp down

on certain industries, essentially by increasing audits and assistance and communicating this

increase, (3) changing the frequency of reporting, (4) using moral suasion by emphasising the

importance of paying taxes, (5) reminding customers about their role in facilitating businesses

not reporting income requesting formal transactions in order to reduce opportunities for

evasion, (6) informing sellers of their relative reporting compared to industry peers and (7)

informing buyers about other buyers willingness to accept cash offers. Exploring these policy

measures relative to a baseline provide the foundation for the experiment.

Results from the experiment indicate that not all treatments are effective to the same degree.

While ATO assistance (1), increasing ATO focus for an industry (2) and moral suasion (4)

significantly increase tax declaration, reminding customers about their role (5), and using

industry peer effects (6) do not have a significantly positive effect. Reducing the frequency of

declarations (3) reduces compliance, but the effect is insignificant. Informing buyers about

peer behaviour (7) does not change their willingness to reject cash offers.

While these results provide a positive outlook for the interventions with a significant effect, it

particularly speaks for using moral suasion to increase compliance, as it may be implemented

at relatively low cost. However, a carrot (offering assistance to meet tax obligations) as well as

a stick (increasing the threat of audits by announcing a focus on a specific industry) approach

may be promising in increasing compliance, especially where there is an evasion opportunity

3 | P a g e

in cash-for-service payments between small businesses and individual customers who may

share a common profit from tax evasion. A stick approach is in particular efficient for those

who are inclined to cash transactions.

4 | P a g e

Contents Executive Summary ................................................................................................................................. 3

Introduction ............................................................................................................................................ 6

Policy Instruments ................................................................................................................................ 10

Experimental Design and Procedures ................................................................................................... 16

Results ................................................................................................................................................... 20

Discussion.............................................................................................................................................. 29

References ............................................................................................................................................ 30

Appendix ............................................................................................................................................... 33

Measurement of the Shadow Economy ........................................................................................... 33

Controlling for nationality of participants in the experiments......................................................... 35

Screenshots ...................................................................................................................................... 36

5 | P a g e

Introduction Although maintaining a high level of tax compliance is central to a well-functioning society,

one crucial challenge to research in this area is that the inherently hidden nature of tax

evasion makes data collection difficult. Increasing self-employed business owners’ tax

compliance (particularly of trade-specific service providers) remains an ongoing challenge

for tax authorities. Cash transactions are particularly problematic from a compliance point of

view when services are paid for on the spot. Such exchanges are difficult to audit, allowing

businesses to conceal income. The building, renovating, and repairing services are

particularly prone to unreported cash transactions. Data from several OECD countries

indicate that the building, renovating, or repairing sector is responsible for almost 50% of the

illicit work (Schneider and Enste 2002).

In Figure 1 we report the level of shadow economy activity as a % of the official

GDP over time. This is calculated using different methods that are explained in the Appendix.

We included three different proxies as the literature has demonstrated that different

calculation methods lead to different results (Schneider and Enste 2002). Such differences are

also visible in Figure 1. In more recent years we observe a rate between 13 and 16%.

Compared to an average of the European countries (for which data are available), Australia

consistently reports values below that of Europe, although there is substantial heterogeneity

within European countries (see Figure 2).

Figure 1: Shadow Economy in Australia over Time

12

14

16

18

Sha

dow

Eco

nom

y (%

of o

ffici

al G

DP)

in A

US

1980 1985 1990 1995 2000 2005 2010Year

Alm & Embaye (2013) Elgin & Öztunali (2012) Schneider et al. (2010)

6 | P a g e

Figure 2: Shadow Economy of Australia Relative to Europe

Note: European countries: Luxembourg, Norway, Switzerland, Denmark, Sweden, Ireland,

Netherlands, Austria, Iceland, Finland, Belgium, Germany, United Kingdom, France, Italy, Spain, and

Portugal.

Enforcement certainly matters in enhancing compliance, as it affects the financial

considerations that motivate – at least in part – an individual’s compliance choice. Yet,

evidence has shown that individuals do not pay taxes only because of the economic

consequences of punishment (Torgler 2002, 2007). Recently, the understanding of individual

choice processes has been expanded by introducing aspects of behaviour or motivation that

can be classified under the general rubric of “behavioural economics”. This is broadly

defined as an approach that uses methods and evidence from a variety of social sciences to

121416182022

Alm

& E

mba

ye (2

013)

1985 1990 1995 2000 2005Year

121416182022

Elg

in &

Özt

unal

i (20

12)

1980 1985 1990 1995 2000 2005 2010Year

121416182022

Sch

neid

er e

t al.

(201

0)

1999 2001 2003 2005 2007Year

Australia European countries averageEuropean countries 95% CI

7 | P a g e

inform the analysis of individual and group decision making. It is therefore valuable to

explore potential instruments by applying a more complete understanding of individual (and

group) decisions, and one that is more consistent with empirical evidence. Evidence in the

area of behavioural economics has demonstrated that many individuals are motivated by

social norms and intrinsic motivation, and that individuals are capable of learning social

norms (Ostrom 2005, Torgler 2007). Research in the area of tax compliance has convincingly

argued that successful tax collection is not only the exercise of power (Alm et al. 2010,

Kirchler 2007, Torgler 2007): tax compliance, like all human behaviour and institutions, is

comprised of a mixture of “love” and “fear” (Boulding 1981). More recently, however, both

researchers and tax administrations have placed more emphasis on integrating the “love”

aspect, especially given that citizens’ consent to pay taxes reflects identification with the

taxing authority’s objectives (Boulding 1981). In this report we therefore use laboratory

experiments to test a set of different instruments so that we may understand compliance

behaviour in a situation where cash transactions are possible. We focus on (1) offering

assistance to increase compliance by the Australian Tax Office (ATO), (2) conducting a

clamp down on certain industries, essentially by increasing audits and communicating this

increase, (3) changing the frequency of reporting, (4) using moral suasion by emphasising the

importance of paying taxes, (5) reminding customers about their role in requesting formal

transactions in order to reduce evasion opportunities and (6) informing sellers of their relative

reporting compared to industry peers. Exploring these policies relative to a baseline provide

the foundation for the experiment.

Because collecting primary data on tax evasion has proven difficult, laboratory experiments

have been an essential tool of thought and exploration in the tax compliance area as

researchers strive to generate their own data (for an overview, see Andreoni, Erard and

Feinstein 1998, Alm 1999, Torgler 2002, Alm 2012). The beauty of this approach lies in the

ability to experimentally test researcher interests while isolating the effect under exploration.

Internal validity is also high because other scientists can replicate the experimental

conditions. Thus, it is not surprising that we observe an increasing number of laboratory

experiments since the 1990s (see Figure 3). Most of these tax compliance experiments use

students as participants, which raises concerns over whether students have enough experience

in this setting. In other words, are students a satisfactory sample for studies on tax behaviour?

Obviously, results should be interpreted with caution (see Webley et al. 1991). Often they

8 | P a g e

correspond to a subject pool with a higher education and a higher IQ than the average citizen.

They often come from families with a higher income than the average and their age ranges

are limited (Fehr et al. 2003). However, there is evidence that students’ responses are not

different from those of other subjects (see, e.g., Baldry 1987). Moreover, so far there is no

reason to think that the cognitive processes of students are different from other subject pools

(Alm 1999). On the other hand, in their cross-country experiments in the Netherlands and

Albania, Gërxhani and Schram (2006) show subject pool differences. Thus, results so far are

not conclusive enough to make a determination either way. One avenue for future research is

to re-test the experiments similar to the ones presented in this report on a larger scale in the

‘real world’, i.e. using small business owners as participants. Such a validation will be useful

to help the ATO to affirm (or not) this methodology research and intervention development.

Figure 3: Yearly Number of Published Laboratory Tax Compliance Experiments

Source: Torgler (2016).

9 | P a g e

Before introducing the experimental design and the results we provide a short

overview of results in the area of tax compliance that are closely related to the different

treatments conducted.

Policy Instruments

Moral Suasion

Economists are generally skeptical about the effects of moral suasion, particularly in the long

term or competitive environments (for a discussion see Torgler 2004a). On the other hand,

social psychologists have demonstrated the power of moral suasion or moral appeals (see,

e.g., Cialdini’s 2007 seminal work on persuasion). Research in marketing especially relies

heavily on persuasion as a tool to influence human nature, as marketing aims at forming and

changing attitudes and behaviours (Torgler 2013). Moral suasion is frequently used in the

public and private sector. In the 70s, for example, petrol enterprises such as Shell ran large

marketing campaigns for using unleaded gasoline despite higher prices. However, after a

short time a drop in sales of unleaded petrol was observed. Shell’s unleaded gasoline “Shell

of the Future” reached only 5 percent of sales (Baumol and Oates 1979). Moral suasion has

shown to be quite efficient in a state of emergency, a disaster or in voluntary contributions

such as blood donation (Baumol and Oates 1979). Tax amnesties have used moral suasion to

promote participation. A successful example was India’s tax amnesty in 1997, which relied

heavily on moral suasion (Torgler 2004a). Moral suasion may help to frame tax compliance

as a positive act (Hasseldine 2000). Over the past 10 years, researchers have gathered

substantial knowledge about how moral sentiments or social norms of compliance (tax

morale) shape tax compliance and the factors that influence such sentiments (for an overview,

see, e.g., Kirchler 2007, Torgler 2007). These observations indicate that social norms are

crucial to understanding why people comply with the law (see also Alm, Martinez-Vazquez

and Torgler 2010). Less evidence is available, however, on how moral suasion or moral

appeals shape tax compliance (for a discussion see Torgler 2004a, 2013). One of the first tax

compliance experiments was carried out by Schwartz and Orleans (1967). Their field

experiment was designed to determine the effects of moral appeals and threats of punishment

10 | P a g e

on behavioural compliance with the tax laws. They found that moral appeals had a much

stronger influence than threats of punishment. These findings were important as they focused

attention on different potential compliance factors. However, since then, little work has been

done to analyse the relevance of moral appeals. In line with Schwartz and Orleans, McGraw

and Scholz (1991) analyse the effects of moral suasion on tax compliance. In the experiment,

the participants were shown a video appealing to social responsibility, but researchers did not

observe a larger increase in income reporting compared with that reported by the control

group. One field experiment by Blumenthal, Christian, and Slemrod (2001), in collaboration

with the Minnesota Department of Revenue, examines the impact of moral persuasion on

voluntary income tax compliance. The study focused on whether taxpayers who are subject to

moral appeals change their reports to a greater extent than taxpayers who are not. They

provide evidence that those receiving the moral suasion treatment increase their reported

income on average by $220 more than the control group (0.8% of the federal taxable income

based on the average between the control and treatment groups), although the coefficient is

not statistically significant. They also show that people with greater opportunities to evade or

avoid taxes (e.g., the self-employed) are less susceptible to normative appeals. The minimal

effect of moral suasion on taxpayers’ compliance behaviour is also demonstrated by Torgler’s

(2004) exploration of moral suasion’s influence on the timely completion of 2001 tax forms

and timely paying of taxes. Torgler (2013) explores individuals’ under-declaration of income

and wealth and over-deduction of tax credits. The letter, similar to the moral suasion used in

the respective treatment in the experiment presented in this report, was supposed to trigger

moral suasion but served also to operationalize social identity and (mutual) trust, which have

been shown to influence tax compliance (for an overview, see Kirchler 2007). However,

moral suasion has barely any effect on tax compliance, even when the field experiment

happened at the local level where moral suasion might be more effective than at the national

level. Thus, the mixed results obtained thus far indicate the necessity of generating further

evidence at the lab and field experimental level.

Assistance

Providing assistance to the taxpayers encourages identification with the task and with the

entity offering assistance, which in this case is the tax administration. It can promote

11 | P a g e

reciprocity through increased obligation. Humans tend to repay what another has provided,

feeling internally obligated to reciprocate favours (Boulding 1981). The web of indebtedness

is also a deeply rooted adaptive mechanism in human nature (Cialdini 2007). Such reciprocity

can contribute to the development of integrative structures between taxpayers and the tax

administration. Failing to comply in future would trigger internal discomfort and

psychological costs (Erard and Feinstein 1994). Mazar et al. (2008, p. 634) stress that if “a

person fails to comply with his or her internal standards for honesty, he or she will need to

negatively update his or her self-concept, which is aversive. Conversely, if a person com-

plies with his or her internal standards, he or she avoids such negative updating and maintains

his or her positive selfview in terms of being an honest person” (p. 634). As a consequence,

Mazar et al. (2008) maintain that people will comply with their internal standards even if

compliance necessitates an investment of effort and sacrifice. Previous research has found

neuroscientific evidence for this principle (Dulleck et al. 2016). Using data from the US and

Turkey, Torgler, Demir, Macintyre and Schaffner (2008) analyse interactions with the tax

administration, finding that positive attitudes towards the tax authority (e.g., how taxpayers

rated tax administrations’ job, their honesty and fairness, and their helping and information

behaviour) significantly increase tax morale. A respectful and fair treatment of taxpayers

induces respect for the tax system and therefore promotes co-operation (Smith 1992).

Torgler and Murphy (2004) report a strong increase of tax morale in Australia during

the 1980s and 1990s. The 1980s saw public administration face growing state and public

demands to become more market-focused, service oriented, open and efficient (Job and

Honaker 2003, Hughes 1994). In response, the ATO adopted a new organizational structure

designed to make them more efficient and customer-focused. Instead of focusing so much on

compliance management, risk control, or structuring the application of enforcement

discretion, the ATO slowly became more focused on service, customers, quality,

transparency and process improvement (see Job and Honaker 2003). In fact, the ATO was

amongst the first tax administrations in the world to implement a new client-based

organizational structure (the client based model is where staff members are assigned to units

that focus on specific groups of customers; for example, salary and wage earners, small

business income taxpayers, and large business income tax payers). One of the advantages of

such a client-based structure is that it allows tax administrations to better match their

enforcement and educational programs to the compliance patterns of different groups

12 | P a g e

(Vehorn and Brondolo 1999). According to Verhorn and Brondolo (1999, p. 505), such a

system has “the potential for delivering higher quality service to taxpayers and achieving

high levels of compliance“. If taxpayers feel that such an approach is likely to achieve better

compliance levels among other taxpayers, then this may go on to influence their own tax

morale. Such a client-focused approach is also likely to increase trust among taxpayers, as

taxpayers are more likely to feel that their needs are being considered in the regulatory

process.

Another example of successful tax administration reform is found in Japan. The

country had to be reorganized after World War II and struggled with high tax delinquency

rates (for a discussion see Torgler 2004b). Compliance improved through significant efforts

of the tax agency to provide the public with tax information via television, radio, internet and

newspapers, distributing pamphlets and booklets. In addition, the administration organized a

“Know-Your-Tax” session every November, with round-table discussions and essay

competitions for students about taxation (Okada 2002). Japan now has relatively high tax

morale values compared with other countries in the region (Torgler 2004b).

Peer Effects

In the early 1980s, the Australian government faced numerous complaints about the existing

income tax system (Smith 1993). The prevailing perception amongst the public was that

many were not meeting their taxation obligations and it was clear that taxpayer compliance

had eroded. “Tax evasion was also contributing to public resentment towards the existing and

highly visible income tax burden” (Smith 1993, p. 111). For example, during the late 1970s,

the scandalous “bottom of the harbour” schemes were being widely publicized. Here,

company profits were being stripped before they could be taxed and the records conveniently

lost (Levi 1988). The more widespread the knowledge that others were not paying their share,

the stronger the perception that non-compliance had increased. As a consequence, taxation

reform was a prominent issue for taxpayers during the mid to late 1980s, and the reforms

“substantially improved the fairness of the tax system at the time” (Mathews 1985, p. 424).

There is substantial evidence that peer effects matter in the area of tax compliance (for

an overview see Frey and Torgler 2007). Two studies in the 80s ran experiments to

13 | P a g e

investigate social comparisons, with mixed results (Spicer and Becker 1980, Webley, Morris

and Amstutz (1985). However, these two experiments were designed to analyze the causal

relationship between inequity and tax evasion. Frey and Torgler (2007) look at survey data

from 30 West and East European countries, providing support for the idea that an individual

taxpayer is strongly influenced by what he or she perceives to be the behaviour of other

taxpayers. If taxpayers believe tax evasion to be common, their tax morale decreases; if they

believe others to be honest, their tax morale increases. The size of the effect is substantial: on

average, when the extent of tax evasion of others increases by one unit (on a scale from 1 to

4), the percentage of persons reporting a high tax morale falls by 7.4 percentage points. Such

peer effects have been observed in other illegal or non-compliant activities such as

assassinations, hijackings, corruption, kidnappings, serial murders, and littering (Bikhchandi,

Hirshleifer and Welch 1998, Dong, Dulleck, and Torgler 2012, Torgler, Frey and Wilson

2009). Kahan (1998) suggests that the decision to commit crimes in general is highly

interdependent, based on the perceived behaviour of others: “When they perceive that many

of their peers are committing crimes, individuals infer that the odds of escaping punishment

are high and the stigma of criminality is low. To the extent that many persons simultaneously

draw these inferences and act on them, moreover, their perceptions become a self-fulfilling

reality” (p. 394).

Focusing Audits on a Specific Industry

Traditionally, audit rates has been a key variable of exploration (particularly among

economists). Evidence from lab experiments tends to indicate that a higher audit rate leads to

more compliance (Alm 1999, Torgler 2002). A meta-study focusing on lab experiments

indicates that raising the audit rates will increase tax compliance (Blackwell 2010). Witte and

Woodbury (1985) report evidence from the field, analysing data for the period 1977-1986

from the Tax Compliance Measurement Program (TCMP) in the US. Their study finds that a

decline in federal audit rates (from 2.5 percent in 1977 to 1 in 1986) led to a reduction in tax

income reporting of $41 billion. However, there is also experimental evidence that the impact

appears nonlinear (diminishing effects) (Alm, McClelland and Schulze 1992). There is also

field experimental evidence indicating that an increase in audit rates matters. Slemrod,

Blumenthal and Christian (2001) used a controlled field experiment in Minnesota to analyse

14 | P a g e

taxpayer response to an increased probability of audit. 1724 randomly selected taxpayers

were informed by letter that the return they were about to file (state and federal) would be

closely examined. They found that the audit effect varies depending on the income. In the

treatment group, low and middle income taxpayers increased their reported income.

15 | P a g e

Experimental Design and Procedures In order to address the question of tax compliance by small business owners in the presence

of different tax enforcement regimes, an experimental design was chosen which used a

stylised interaction between service providers and customers. More particularly, participants

were told that they would make decisions in a service provider-customer framework and

would interact with other participants in these roles throughout the experiment.

Upon arrival at the laboratory participants were welcomed and it was explained that

they would make decisions throughout the experiment, for which they would be paid (see

Figure A1 in the Appendix). This ensures that participants have an incentive to do what they

like best – if they would like to choose the dishonest declaration, they could do so and reap

the economic benefits from doing so. If they made an honest declaration instead, they would

need to give up some monetary income at the end of the experiment.

After making participants familiar with the rules that would determine their payments

(see Figure A2 in the Appendix), the structure of the experimental interaction was explained

(see Figure A3). They were told that there would be two roles, those of the service providers

and those of the customer. In each round of interaction the customer would need to get a job

done in order not to lose 80 experimental dollars (from the 100 experimental dollars the

customer would receive each period). At the same time the service provider was able to get

this job done and was able to make an offer for getting the job done to the customer.

Participants were told that they could think of it as the relationship between a house owner

and a tradesperson:

“To understand the interaction between service providers and consumers consider the

consumer to be a house owner who needs some job to be done in the house which he cannot

do himself. In real life this could be the repair of your swimming pool, of your hot water

system or the refurbishment of a fence around your house. Whilst these jobs may not need to

be fixed straight away, there is a greater cost of ignoring the problem. The service provider in

turn is someone who can do the job, such as a pool repair person, a plumber or someone

specialised in fixing fences” (see Figure A3).

Furthermore, participants were told that the service providers had to pay taxes on their

income, hence the price paid from customers to service providers. The tax rate was 40% (see

Figure A4). Furthermore, service providers were asked if they wanted to offer a cash option

16 | P a g e

to customers, and customers were asked if they would accept a cash offer (both were asked

independently and the cash offer was only implemented if both had opted for the cash

option). Cash offers implied that the price to be paid by the customer would be 10% lower.

Furthermore, in case the cash option was implemented, the income (hence, the price minus

10%) of the service provider was not automatically taxed. Instead, in this case the service

provider had to declare how much had been received. Hence, in case of cash transactions, it is

possible to evade taxes by under-declaring income. Non-cash transactions were in turn

automatically “declared” correctly. Participants were also informed that money collected as

taxes would not be lost, but paid to a university charity.

In addition, participants were told that after each declaration there was a possibility

that the service provider would be audited. The audit probability (in the baseline treatment)

was set to 10%, and failing the audit had consequences. If a service provider was audited and

the amount declared was lower than the correct amount, the service provider would have to

pay a fine, which corresponded to the underpaid amount plus the difference of the underpaid

taxes.

Participants were then advised that there would be two parts of the experiment (see

Figure A5). In the first part they would be either a service provider or a customer and then

they would switch roles in the second part. Participants were also told that each part had 6

rounds and that in every round a different customer would interact with a different service

provider. Participants then proceeded to the decision rounds. Furthermore, they were advised

that there would potentially be further instructions on later screens and were therefore asked

to read these carefully. In experimental terms, these further instructions represent the

different treatments throughout the experiment. The treatments are outlined in further detail

below. Furthermore, before starting with the decision making, participants had to answer two

control questions to ensure that they understand the game and how they would be paid based

on their decisions. Control questions had to be answered correctly before participants were

able to advance in the experiment. The full instructions showing the buyer and seller

interactions including the auditing are reported in the Appendix (see Figures A6 to A11).

The experiment utilised the baseline as described above, and six further treatments.

Treatments were introduced by further descriptions on participants’ computer screens. The

following lists the treatments used (see also Figures A6.1 to A7.3 in the Appendix):

17 | P a g e

1. The Baseline, participants in this control group received full information about the

experiment as outlined above with an audit probability of 10%.

2. The ATO assist, hence, it was pointed out to participants that they could request

further help on how to comply by asking a research assistant who was available for

compliance questions.

3. The Clamp down, which notified participants that in the following three periods the

industry of service providers had come under special scrutiny, implying a doubling of

the audit probability (relevant for the sellers).

4. The infrequent reporting, meaning that service providers did not have to report their

income every period but rather after 3 periods as one large instalment.

5. The moral suasion, meaning that both customers/buyers and service providers/sellers

were reminded that tax money served a common good and that it would be paid to

charity, namely a food bank at the university (QUT). Therefore, it was pointed out

that paying taxes in this experiment was important from a common good perspective.

6. The peer effects buyer, during which it was pointed out to buyers that accepting cash

offers would provide sellers with the opportunity to evade taxes – and that buyers

could consequently play a part in increasing compliance.

7. The peer effects seller, for which sellers were informed that their declaration was

below/about/above the industry average of declared income based on previous

experiments. While not asking sellers to reconsider their declaration, they had the

option to then adapt the amount declared.

The actual tax declaration screen is shown in Figure 9. Once the experiment was complete

(i.e., after the two rounds of 6 periods each) participants were asked to fill in a post-

experimental questionnaire, which provided further demographic information and self-

reported attitudes of participants. The full list of questions is included in the Appendix (see

Figure A12). Questions relating to tax compliance, demographic questions (gender, age,

nationality) and business ownership were compulsory, while other (e.g. income, religion)

were answered on a voluntary basis.

18 | P a g e

The experiment was conducted in a computer laboratory at Queensland University of

Technology in Queensland, Australia, between June 7 and June 24 2016 with 200 volunteer

participants. Of these 16 were used in a pilot session, and were therefore excluded for later

analysis, resulting in a final number of 184 participants of which 42.93% were female, the

average age was 23 and 47.83% of participants were Australian nationals. The sample of

participants consisted of undergraduate and postgraduate students, and the largest proportion

of participants were undergraduate students (71.67% Bachelors, 19.44% Master and 8.89%

PhD).1 We do control in our result section for these characteristics, including the nationality

of participants.

At the end of the experiment, participants were paid for their decisions. Experimental

payments were first calculated in experimental dollars based on two randomly chosen

interactions and subsequently exchanged into the Australian dollar at a rate of 0.75

experimental dollars = 1 AUD as announced at the start of the experiment. On average,

participants earned (including a show-up fee of 10 Australian dollars for participating in the

experiment) 44.18 Australian dollars (SD = 10.38) throughout the experiment, which lasted

approximately 55 minutes.

1 Due to the restricted time frame the baseline was conducted by itself whilst the remaining six treatments were grouped into a session as follows, ATO assist and clamp down, infrequent and moral suasion, and peer effects buyer and peer effects seller. Of the seven treatments conducted, infrequent and moral suasion have the most participants (54), whilst the baseline has the greatest number of interactions (204). ATO assist and clamp down had 50 participants, followed by peer effects buyer and seller with 46 participants. Due to technical difficulties, the observations of 17 participants in moral suasion were affected, resulting in 142 interactions instead of 150.

19 | P a g e

Results The data generated by experimental decisions were loaded into the STATA statistical

software for analysis. The decision variables of the experiment can be summarised as

follows: The average amount earned was 64 experimental dollars (standard deviation = 14),

the frequency of cash offers was 57% and the average declared amount was 48 experimental

dollars (SD = 30) for service provider decisions. For customers, the 78% of offers were

accepted, and the frequency of customer accepting cash offers was 85%. These numbers

show that there was significant scope for evasion and that cash offers were frequently used

and substantial amounts were evaded. It is important to note that caution is necessary to apply

to absolute levels observed in the laboratory. The absolut values reflect decisions inside the

laboratory and similar levels in the real world are coincidental, ie. levels in the real world

may be substantially different. This methodology focusses on the difference between the

treatments and the control group, the qualitative effects of treatments or interventions have

been shown to be similar in the laboratory and the real world.

Based on the experimental design, the main variable of interest is the amount declared

by service providers (also including the cases where no cash offer was provided). Hence, we

are not just looking at tax evasion but rather the final declared amount, which is of crucial

interest for the tax administration (how much tax money comes in). In our setting, we can

directly focus on the declared amount as we have the same tax rate for all the relevant

treatments and the control group. Figure 10 shows the average amount declared separated by

treatments. As can be seen, the amount declared appears to be higher in some treatments than

in others, whereas all relevant treatments appear to increase compliance. It is clear from

Table 1 that the differences between the control and the relevant treatment groups are

statistically significant (with the exception of the infrequent treatment). Moral suasion and

assistance seemed to have the strongest effect on income declaration. We have excluded from

the figure the peer effect buyer treatment, as this treatment should only have an effect in

terms of increasing the rejection rate of cash offers by the sellers. Thus, such a treatment is

similar to a control group from a seller perspective. To our surprise, this treatment reported a

higher declaration rate. Thus, we added that treatment to the control group in Figure 11. The

treatments of moral suasion, assistance and clamp down still show a statistically significant

difference to the control group.

20 | P a g e

Figure 10: Declared amount in the control group and the treatment groups

Table 1: Differences between the groups

Treatment Mean difference t-statistic N (treatment) N (baseline) Rank-sum

ATO assist 16.47*** -5.18 100 144 -4.72***

Clamp down 16.71*** -4.38 61 144 -3.97***

Infrequent 0.87 -0.26 162 144 -0.58

Moral suasion 16.99*** -5.04 137 144 -4.78***

Peer effects seller 5.5* -1.79 108 144 -1.32

Notes: Results for two-sample mean comparison t-test (column 1 and 2) and non-parametric Wilcoxon rank-sum

test (column 5). Significance is indicated as follows: * indicates 10% significance, ** 5% significance and ***

1% significance.

38.58

55.05 55.29

39.45

55.57

44.09

020

4060

Mea

n de

clar

ed a

mou

nt

Baseli

ne

ATO Ass

ist

Clamp D

own

Infreq

uent

Moral S

uasio

n

Peer E

ffects

Seller

21 | P a g e

Figure 11: Adding the baseline and the peer effects buyer into the control group

Table 2: Differences between the groups changing the control group

Treatment Mean difference t-statistic N (treatment) N (baseline) Rank-sum

ATO assist 10.94*** -3.82 100 250 -3.91***

Clamp down 11.18*** -3.19 61 250 -3.14***

Infrequent -4.66 1.64 162 250 1.15

Moral suasion 11.46*** -3.98 137 250 -4.07***

Peer effects seller -0.03 0.01 108 250 0.73

Notes: Results for two-sample mean comparison t-test (column 1 and 2) and non-parametric Wilcoxon rank-sum

test (column 5). Significance is indicated as follows: * indicates 10% significance, ** 5% significance and ***

1% significance.

Next, we take a closer look whether text introduced in the peer effects buyer treatment

reduces the likelihood of accepting a cash offer. For this, we compare the peer effects buyer

44.11

55.05 55.29

39.45

55.57

44.09

020

4060

Mea

n de

clar

ed a

mou

nt

Baseli

ne &

Pee

r Effe

cts B

uyer

ATO Ass

ist

Clamp D

own

Infreq

uent

Moral S

uasio

n

Peer E

ffects

Seller

22 | P a g e

treatment with all the other treatments including the control group. In Figure 12 we can see

that the introduction of such text has only a small effect that is not statistically significant.

Figure 12: Acceptance of cash transfers

Notes: Difference is not statistically significant. All other treatments and the control group in the group

labelled “All others”.

In the multivariate analysis we first present empirical results adding the peer effects

buyer treatment to the control group (see specifications A to C). We are therefore generating

conservative estimates. In specifications D to E we then exclude the peer effects buyer

treatment from the specifications. We conduct a multivariate analysis, because it could be the

case that the effects visible in the descriptive analysis may be driven by the fact that

individuals were observed repeatedly. Furthermore, as indicated by the design features

described in footnote 1, there may be session effects. For this reason, statistical differences

were tested using multilevel regressions, which are a statistical tool designed to control for

such effects. Table 3 shows the results from this analysis, comparing the amount and

successively introducing further controls, namely demographics (B, E) as well as the price

offered to the consumer (C, F). Additionally, the domestic status of the participant was

0.85 0.84

0.2

.4.6

.81

% A

ccep

ting

cash

offe

r

All others Peer Effects Buyer

23 | P a g e

further controlled for (see appendix Table A1), where associating with an Australian

nationality is statistically significant in increasing the amount declared. As can be seen in the

appendix, the results presented are robust when controlling for this characteristic of

participants. The result is also represented graphically as the difference to the control group

in Figures 13 and 14. As can be seen, ATO assist, clamp down and moral suasion

significantly increase the amount declared. Infrequent reporting leads to (insignificantly)

lower amounts declared and seller peer effects leads to slightly higher (but not statistically

significant) declared income.

Table 3: Multilevel random-effects regressions of determinants of the declared amount

Dep. Var. = Declared amount

A B C D E F

treatments ato assist 10.45** 9.93** 8.15* 15.21*** 16.23*** 14.05*** (2.12) (2.06) (1.77) (2.82) (3.10) (2.80) clamp down 11.14** 10.78** 10.26** 15.88*** 17.13*** 16.10*** (2.14) (2.12) (2.10) (2.81) (3.10) (3.04) infrequent -5.04 -6.98 -6.64 -0.30 -0.90 -1.14 (-1.08) (-1.52) (-1.52) (-0.06) (-0.18) (-0.24) moral suasion 10.27** 9.72** 7.78* 15.05*** 16.05*** 13.73*** (2.18) (2.11) (1.77) (2.90) (3.19) (2.85) peer effects seller

0.59 1.07 3.23 5.29 7.50 8.87*

(0.12) (0.22) (0.69) (0.97) (1.40) (1.74) Further controls

male -7.22** -6.82** -8.42*** -7.90*** (-2.40) (-2.38) (-2.65) (-2.61) age 0.59** 0.55** 0.69** 0.64** (2.15) (2.08) (2.36) (2.30) control for price offered

0.38*** 0.35***

(4.98) (4.19) Constant 44.50*** 42.51*** 19.06** 39.75*** 35.84*** 15.01 N 818 818 818 712 712 712 Prob. > chi^2 0.007 0.000 0.000 0.001 0.000 0.000 Notes: Reference in A, B, and C: baseline + peer effects buyer treatment. D, E,F only baseline. Standard errors are clustered by individual and session. Significance is indicated as follows: * indicates 10% significance, ** 5% significance and *** 1% significance.

24 | P a g e

Figure 13: Estimated differences to the joint control group

Figure 14: Estimated differences to baseline

10.2111.21

-6.01

9.66

0.37

-8-4

04

812

Mea

n E

stim

ated

Diff

eren

ce to

Con

trol G

roup

ATO Ass

ist

Clamp D

own

Infreq

uent

Moral S

uasio

n

Peer E

ffects

Seller

Note: Control group is equal to Baseline & Peer Effect Buyer.

14.8015.87

-1.36

14.32

4.92

-40

48

1216

Mea

n E

stim

ated

Diff

eren

ce to

Con

trol G

roup

ATO Ass

ist

Clamp D

own

Infreq

uent

Moral S

uasio

n

Peer E

ffects

Seller

Note: Control group is equal to Baseline. Peer Effect Buyer is excluded.

25 | P a g e

Figure 15: Tax compliance with the mixed reference group

Figure 16: Tax compliance with the baseline

0.52

0.61

0.82

0.44

0.62 0.61

0.2

.4.6

.8M

ean

Com

plia

nce

Rat

io

Baseli

ne &

Pee

r Effe

cts B

uyer

ATO Ass

ist

Clamp D

own

Infreq

uent

Moral S

uasio

n

Peer E

ffects

Seller

Note: Control group is equal to Baseline & Peer Effect Buyer.

0.48

0.61

0.82

0.44

0.62 0.61

0.2

.4.6

.8M

ean

Com

plia

nce

Rat

io

Baseli

ne

ATO Ass

ist

Clamp D

own

Infreq

uent

Moral S

uasio

n

Peer E

ffects

Seller

Note: Control group is equal to Baseline. Excludes Peer Effect Buyer

26 | P a g e

Next, we take a closer look at tax compliance of those sellers who suggested a cash transaction that

was accepted by the buyers. Overall, such sellers might be more inclined to evade taxes and

therefore we would predict that softer factors such as assistance or moral suasion would work less

strong. Figures 15 and 16 and Table 4 support such predictions. The clamp down treatment has by

far the strongest effect (coefficient is always statistically significant), followed by moral suasion.

27 | P a g e

Table 4: Tax compliance for those sellers who used cash transactions Dep. Var. = Tax Compliance Ratio

G H I J K L

treatments ato assist 0.07 0.07 0.11 0.10 0.15 0.18 (0.49) (0.56) (0.89) (0.65) (0.96) (1.27) clamp down 0.33** 0.32** 0.36*** 0.36** 0.39** 0.43*** (2.27) (2.26) (2.64) (2.23) (2.45) (2.80) infrequent -0.14 -0.16 -0.17 -0.11 -0.09 -0.11 (-1.03) (-1.16) (-1.34) (-0.71) (-0.59) (-0.74) moral suasion 0.18 0.16 0.21* 0.21 0.23 0.28** (1.39) (1.29) (1.76) (1.44) (1.62) (2.04) peer effects seller

0.06 0.08 0.06 0.09 0.14 0.13

(0.43) (0.57) (0.45) (0.59) (0.95) (0.88) Further controls

male -0.19** -0.17** -0.19** -0.18** (-2.21) (-2.17) (-2.02) (-2.06) age 0.00 0.00 0.01 0.01 (0.63) (0.54) (0.71) (0.68) control for price offered

-0.01*** -0.01***

(-3.87) (-3.72) Constant 0.58*** 0.76*** 1.14*** 0.55*** 0.67*** 1.06*** N 401 401 401 350 350 350 Prob. > chi^2 0.090 0.036 0.000 0.107 0.053 0.000 Multilevel random-effects regressions of determinants of the declared amount. Standard errors are clustered by individual and session. Significance is indicated as follows: * indicates 10% significance, ** 5% significance and *** 1% significance. J, K, L exclude peer effects buyer completely.

28 | P a g e

Discussion and Recommendations This is one of the first laboratory experiments to investigate the implication of different

policies on income declaration in an environment where cash transactions are possible. Thus,

the results contribute to the literature on the shadow economy that mostly relies on macro

data to derive policy implications. Most of the studies on tax compliance have focused on

personal income tax. The idea of introducing a laboratory study in a business setting has

received very little attention. This is a surprising taking into account the economic

importance of the business sector and the high rates of illicit work or frequent use of cash

transactions in the building, renovating and repairing sector. The study also provides insights

as to how different policies work in an environment where we still do not have enough

insights as to the robustness of such instruments. Contrary to several field experiments we

observe that moral suasion is a very powerful tool in increasing tax revenues. Moreover, we

observe that providing assistance is also crucial for maintaining a high level of compliance

and cooperation. This can be seen as evidence that perceived procedural fairness is a key

factor in guaranteeing sustainable compliance. The results in the clamp down treatment also

indicate that deterrence matters. However, from a comparative point of view, assistance and

moral suasion seemed to have a stronger effect on tax revenues. Thus, taxpayers are not only

driven by extrinsic motivation. It is important to maintain a high level of intrinsic motivation

that can be encouraged with moral appeals and assistance. However, for those who are

inclined to cash transactions, deterrence seemed to have the strongest influence on tax

compliance.

This report has identified three treatments which show good potential for the ATO. We

recommend to extend this study to test an additional set of four interventions in a comparable

way. Furthermore, we recommend to test the most promising interventions – from these as

well as the additional treatments – together with a baseline for testing on a larger scale with

members of the relevant community in the controlled environment of a computer based

experiments. This could then inform the design of a randomized controlled trial in the

community.

29 | P a g e

References Alm, J. (1999). Tax Compliance and Administration, in: W. B. Hildreth, J. A. Richardson (eds.),

Handbook on Taxation. New York: Marcel Dekker: 741-768.

Alm, J. (2012). Measuring, explaining, and controlling tax evasion: lessons from theory, experiments,

and field studies, International Tax and Public Finance. 19: 54-77.

Alm, J. and A. Embaye (2013). Using Dynamic Panel Methods to Estimate Shadow Economies

Around the World, 1984-2006, Public Finance Review. 41: 510-543.

Alm, J., J. Martinez-Vazquez, and B. Torgler (eds.) (2010). Developing Alternative Frameworks for

Explaining Tax Compliance. London: Routledge.

Alm, J., G. H. McClelland and W. D. Schulze (1992). Why Do People Pay Taxes? Journal of Public

Economics. 48: 21-48.

Andreoni, J., B. Erard and J. Feinstein (1998). Tax Compliance, Journal of Economic Literature. 36:

818-860.

Baldry, J. C. (1987). Income Tax Evasion and the Tax Schedule: Some Experimental Results, Public

Finance. 42: 357-383.

Baumol, W. J. and W. E. Oates (1979). Economics, Environmental Policy, and the Quality of Life.

Englewood Cliffs: Prentice-Hall.

Bikhchandani, S., D. Hirshleifer, and I. Welch (1998). Learning from the behavior of others:

Conformity, fads, and informational cascades. Journal of Economic Perspectives. 12: 151-170.

Blackwell, C. (2010). A meta-analysis of incentive effects in tax compliance experiments, in: J. Alm,

J. Martinez-Vazquez, and B. Torgler (Eds.), Developing Alternative Frameworks Explaining

Tax Compliance, eds., Routledge, London.

Blumenthal, M., Christian, C., Slemrod, J. (2001) Do Normative Appeals Affect Tax Compliance?

Evidence from a Controlled Experiment in Minnesota. National Tax Journal. 54: 125-138.

Boulding, K. E. (1981). A Preface to Grant Economics: The Economy of Love and Fear. Praeger.

Cialdini, R. B. (2007). Influence: The Psychology of Persuasion. New York: HarperCollins.

Dong, B., U. Dulleck, and B. Torgler (2012). Conditional Corruption, Journal of Economic

Psychology. 33: 609-627.

Dulleck, U., J. Fooken, C. Newton, A. Ristl, M. Schaffner, and B. Torgler (2016). Tax Compliance

and Psychic Costs: Behavioral Experimental Evidence Using a Physiological Marker, Journal

of Public Economics. 134: 9-18.

Elgin, C. and O. Öztunali (2013). Shadow Economies Around the World: Model Based Estimates,

Bogazici University, Department of Economics, WP 2012/05.

30 | P a g e

Fehr, E., U. Fischbacher, B. von Rosenbladt, J. Schupp and G. G. Wagner (2003). A Nation-Wide

Laboratory. Examing Trust and Trustworthiness by Integrating Behavioral Experiments into

Representative Surveys, CESifo Working Paper No. 866, February 2003.

Frey, B. S. and B. Torgler (2007). Tax Morale and Conditional Cooperation, Journal of Comparative

Economics. 35: 136-159.

Fuest, C. and N. Riedle (2009). Tax Evasion, Tax Avoidance and Tax Expenditures in Developing

Countries: A Review of the Literature, Report prepared for the UK Department for

International Development (DFID), June 19.

Gërxhani, K, and A. Schram (2006). Tax evasion and income source: A comparative experimental

study, Journal of Economic Psychology. 27: 402-422.

Hasseldine, J. (2000) Using Persuasive Communications to Increase Tax Compliance: What

Experimental Research Has (and Has Not) Told Us. Australian Tax Forum. 15: 227-224.

Hughes, O. E. (1994). Public Management and Administration. New York: St. Martin’s Press.

Job, J. and D. Honaker (2003). Short-term experience with responsive regulation in the Australian

Taxation Office, in: V. Braithwaite (Ed.), Taxing Democracy: Understanding avoidance and

evasion (pp. 111-130). Aldershot: Ashgate.

Kahan, D. M. (1998). Social meaning and the economic analysis of crime. Journal of Legal Studies,

27, 609-622.

Kirchler, E. (2007). The Economic Psychology of Tax Behaviour. Cambridge: Cambridge Uni-versity

Press.

Mathews, R. L. (1985). Some reflections on the 1985 Tax Reforms. Australian Tax Forum, 2(4).

Mazar, N., O. Amir, and D. Ariely (2008). The Dishonesty of Honest People: A Theory of Self-

Concept Maintenance, Journal of Marketing Research. XLV: 633-644.

McGraw, K. and J. T. Scholz (1991). Appeals to Civic Virtue Versus Attention to Self-Interest:

Effects on Tax Compliance, Law and Society Review. 25: 471-498.

La Porta, R., F. Lopez de Silanes, A. Shleifer and R. Vishny (1999). The Quality of Government,

Journal of Law, Economics and Organization. 15: 222-279.

Okada, Y. (2002). The Japanese Way, paper presented at the conference New Challenges in Tax

Compliance: Japan’s Experience and its Significance for Latin America, June 11-12.

Ostrom, E. (2005). Policies that Crowd out Reciprocity and Collective Action, in: H. Gintis, S.

Bowles, R. Boyd, and E. Fehr (Eds.), Moral Sentiments and Material Inter-ests. Cambridge,

Mass.: MIT Press, pp. 253-275.

Schneider, F., and D. H Enste (2002). The Shadow Economy. An International Survey. Cambridge:

Cambridge University Press.

Schneider, F. , A. Buehn, and C. E. Montenegro (2010). New Estimates for the Shadow Economies all

over the World, International Economic Journal. 24: 443 – 461.

31 | P a g e

Schwartz, R. and S. Orleans (1967). On Legal Sanctions. University of Chicago Law Review. 34: 282-

300.

Slemrod, J., Blumenthal, M. Christian, C. (2001) Taxpayer Response to an Increased Probability of

Audit: Evidence from a Controlled Experiment in Minnesota. Journal of Public Economics. 79:

455-483.

Smith, K. W. (1992). Reciprocity and Fairness: Positive Incentives for Tax Compliance, in: J.

Slemrod (ed.), Why People Pay Taxes. Tax Compliance and Enforcement. Ann Arbor:

University of Michigan Press: 223-258.

Smith, J. P. (1993). Taxing Popularity: The Story of Taxation in Australia. Federalism Research

Centre, The Australian National University.

Torgler, B. (2002). Speaking to Theorists and Searching for Facts: Tax Morale and Tax Compliance

in Experiments, Journal of Economic Surveys. 16: 657-684.

Torgler, B. (2004a). Moral Suasion: An Alternative Tax Policy Strategy? Evidence from a Controlled

Field Experiment in Switzerland, Economics of Governance. 5: 235-253.

Torgler, B. (2004b). Tax Morale in Asian Countries, Journal of Asian Economics. 15: 237–266.

Torgler, B. (2007). Tax Compliance and Tax Morale: A Theoretical and Empirical Analysis.

Cheltenham, UK: Edward Elgar.

Torgler, B. (2013). A Field Experiment in Moral Suasion and Tax Compliance Focusing on

Underdeclaration and Overdeduction, FinanzArchiv: Public Finance Analysis. 69: 393-411.

Torgler, B. (2016). Tax Compliance and Data: What Is Available and What Is Needed, forthcoming

Australian Economic Review.

Torgler, B. and K. Murphy (2004). Tax Morale in Australia: What Factors Shape It and Has It

Changed Over Time?, Journal of Australian Taxation. 7: 298-335.

Torgler, B., B. S. Frey, and C. Wilson. 2009. Environmental and pro-social norms: Evidence on

littering. B. E. Journal of Economic Analysis & Policy 9(1), Article 18.

Torgler, B., I. C. Demir, A. Macintyre and M. Schaffner (2008). Causes and Consequences of Tax

Morale: An Empirical Investigation, Economic Analysis & Policy. 38: 313-339.

Verhorn, C. L. and J. Brondolo (1999). Organizational options for tax administration. Bulletin for

International Fiscal Documentation. 53: 499-512.

Webley, P., H. Robben, H. Elffers and D. Hessing (1991). Tax Evasion: An Experimental Approach.

Cambridge: Cambridge University Press.

32 | P a g e

Appendix

Measurement of the Shadow Economy

Elgin and Öztunali (2012) use the most classical economics approach, modeling a two-sector

dynamic general equilibrium economy to explicitly estimate the size of the shadow economy.

In essence, their approach assumes that the economy of a specific country consists of a

formal and an informal sector. Since the economy is assumed to be in equilibrium,

perturbations to the economic indicators of that economy then follow the real business cycle.

Using a set of parameters for this expected movement and relying on the fact that most of the

indicators for the formal sector are known; they are able to use historical data to accurately

calibrate the model. Thus, they directly calculate the implied size of the shadow economy.

The approach certainly has its advantages: it is based on a strong micro-foundation and does

not suffer from a heavy reliance on the selection of econometric methods (as is the case with

the other methods). Furthermore, this method produces a direct estimate of the size of the

shadow economy, while the other two methods (discussed below) require additional

assumptions in order to work out the size of the shadow economy based on measures of

relative size. However, the reliability of this approach depends to a great degree on the

underlying model being correct and does not account well for unobserved influences on the

shadow economy.

Alm and Embaye (2013) propose a more empirically based method. The basic

premise is that the shadow economy is a key influence on the demand for currency, hence

driving to some extent the currency demand of a country. The method then calculates the

currency demand impact for a set of variables that are known to influence the demand for

cash, such as the interest rate and level of urbanization. Other indicators for the extent of the

shadow economy are also included; for example, the tax rate and the level of enforcement in

a country. The resulting estimates then allow calculation of the currency demand for the

formal economy in the absence of drivers for the informal economy. The proportional size of

the informal economy is backed out using the relative difference between this formal

economy estimate and the estimate for the actual cash demand in the economy (accounting

33 | P a g e

for all factors). However, the key challenge in these models is to properly take into

consideration the interdependencies between the various drivers of currency demand in the

formal and informal economy across time and between different countries. Alm and Embaye

(2013) account for this factor by using a dynamic panel method, namely a GMM estimator,

which is a state-of-the-art methodology to account for the endogeneity problems outlined

before. Moreover, the advantage of this method is that it uses a limited set of variables, all of

which are available for most countries for a long time period, giving credence to the estimates

produced.

Finally, Schneider et al. (2010) use a similarly empirical, but more methodologically

comprehensive approach to work out the size of the informal economy. Recognizing the

limitations of working with any single indicator, they use a Multiple Indicators Multiple

Causes (MIMIC) model, incorporating a number of different approaches into a unified model.

Technically, their approach employs a Structural Equation Model (SEM) with one latent

variable – the size of the shadow economy. The general idea of a latent variable model is that

it allows indirect insights into the behaviour of a hidden variable. If you have no reliable

single indicator for a hidden entity, but you do have information about the movement of a

number of variables in relation to your variable of interest, you can understand the movement

of the hidden/latent variable by carefully accounting for the covariance of the indicator

variables. This approach is much less restrictive in terms of explicitly modeling the shadow

economy compared with the first two methods discussed, as it statistically exploits an artefact

of the co-movement of variables. However this freedom comes at a cost: the approach can

neither claim the rigorous micro-foundation of the Elgin and Öztunali (2012) approach, nor

can it explicitly control for the endogeneity issues addressed by Alm and Embaye (2013).

Moreover, the MIMIC DYMIMIC approach used by Schneider et al. (2010) is sensitive to

sample changes and adjustments in the set of indicator and cause variables (Fuest and Riedel

2009). As La Porta and Shleifer (2008) point out, the measure “is only as good as the model

that supports it” (p. 282).

34 | P a g e

Controlling for nationality of participants in the experiments Given the relatively large share of non-domestic participants in the experiment, the following regression result show that the results presented in the report are robust, if not stronger, if one controls for this characteristic.

Table A1: Multilevel random-effects regressions of determinants of the declared amount Notes: Reference in M and N: baseline + peer effects buyer treatment. O and P are only baseline. Standard errors are clustered by individual and session. Significance is indicated as follows: * indicates 10% significance, ** 5% significance and *** 1% significance.

Dep. Var. = Declared amount

M N O P

treatments ato assist 10.11** 8.30* 16.08*** 13.75*** (2.14) (1.84) (3.14) (2.82) clamp down 12.12** 11.68** 18.10*** 17.12*** (2.41) (2.44) (3.33) (3.31) infrequent -5.53 -5.09 0.20 0.07 (-1.22) (-1.19) (0.04) (0.01) moral suasion 9.44** 7.43* 15.40*** 12.86*** (2.09) (1.74) (3.12) (2.75) peer effects seller

0.02 2.18 6.01 7.30

(0.00) (0.47) (1.14) (1.47) Further controls

male -8.08*** -7.73*** -8.96*** -8.48*** (-2.72) (-2.75) (-2.87) (-2.87) age 0.88*** 0.85*** 0.99*** 0.98*** (3.02) (3.07) (3.20) (3.33) domestic 8.56*** 9.05*** 8.68** 9.63*** (2.68) (2.99) (2.55) (2.99) control for price offered

0.39*** 0.37***

(5.16) (4.50) Constant 32.76*** 8.11 25.39*** 2.15 N 818 818 712 712 Prob. > chi^2 0.000 0.000 0.000 0.000

35 | P a g e

Screenshots The following pages provide a full set of screenshots for one experimental condition (treatment) of conducted experiments.

Figure A1: Welcome screen

36 | P a g e

Figure A2: Experimental payment

37 | P a g e

Figure A3: Structure of the experimental interaction

38 | P a g e

Figure A4: Cash offer

39 | P a g e

Figure A5: Roles

40 | P a g e

Figure A6.1: Assistance Treatment

Figure A6.2: Clamp Down Treatment

41 | P a g e

Figure A6.3: Moral Suasion

Figure A6.4: Buyer Peer Effect

42 | P a g e

Figure A7.1: Seller’s offer

43 | P a g e

Figure A7.2: Seller’s cash offer

44 | P a g e

Figure A7.3: Seller Peer Effect Treatment

45 | P a g e

Figure A8: Tax Declaration

46 | P a g e

Figure A9: Auditing the seller

47 | P a g e

Figure A10: Buyer’s offer acceptance screen

48 | P a g e

Figure A11: Buyer’s response to a cash offer

49 | P a g e

Figure A12: Buyer’s acceptance of a cash offer

50 | P a g e

Figure A12: Survey

51 | P a g e

52 | P a g e

53 | P a g e

54 | P a g e

55 | P a g e

56 | P a g e

57 | P a g e

58 | P a g e

59 | P a g e

60 | P a g e