1. INTRODUCTION

There is a rich literature that studies the effects of diverse interventions to reduce tax evasion. These interventions are usually complex and involve behavioral or punitive strategies, or a combination of both. This paper studies a rather simple alternative that has not been studied in detail in the literature: reducing the capacity of taxpayers to refile their taxes an unlimited number of times. Between the years of 2010 and 2012, the Ecuadorian tax authority allowed for unlimited tax refiles with an interesting, in hindsight mistaken for tax compliance, feature. A misreporting person could leave the list of “evaders” once she or he refiled their taxes correctly, but the option to refile yet again with a beneficially incorrect amount remained open. The study by ^{Sánchez (2022)} found empirical evidence that notifications of incorrect reporting during the time of this policy led to increased levels of reported taxes, but had no effect on the total amount of tax dollars collected. While this empirical finding makes sense in hindsight, there has been no formal behavioral theoretical analysis that illuminates the conditions under which the ability to do unlimited refilings -i.e. as many times as desired without going through a formal process that verifies that the refile is legally necessary- can lead to adverse for tax compliance outcomes.

This paper, therefore, building upon ^{Sánchez (2022)}, formally models the effect of allowing unlimited tax refiles on total taxes collected in the context of the tax control policies applied in Ecuador between 2010 and 2012 -it should be highlighted it limited refiling since 2013 by requiring that after the first refile, a formal petition must be submitted and reviewed by a tax authority. This analysis formally clarifies why, although tax notifications sent to taxpayers who have under-reported the income tax advance (ITA) in Ecuador increased reported taxes, the notifications did not affect the total amount of collected taxes. The model also supports the hypothesis that some taxpayers initially refiled their taxes to correct their “miscalculations” and then exit the list of tax evaders, but later refiled again to evade taxes. The support is stronger when evading the ITA implies reducing the tax liability.

More specifically, the investigation finds, first, that, if unlimited refilings are possible and taxpayers are exclusively selfish -i.e., only care about their individual payoffs-, then their best decision -i.e., that maximizes expected payoffs- is to under-report their ITA even if tax authorities never miss their under-reporting¹. A key implication of this first finding of the paper (which explains its title) is a call for policymakers to carefully think, in advance of applying some policy, if it would actually change the incentives of selfish economic agents in the desired ways.

In order to explore possible solutions, the model is then extended, in two steps. First, remaining in the context of selfish taxpayers, the effects of banning unlimited refilings are examined, finding that such banning is necessary but not sufficient to achieve that selfish taxpayers report the truth and that having a proficient tax authority which often catches under-reporters is key to successfully curb tax evasion in this scenario. Second, the effects of having taxpayers who to some degree exhibit social preferences -i.e., to some extent care about the effects of paying taxes for society at large²- on tax compliance are examined, both in the case in which unlimited refilings are allowed and in the case in which they are banned, finding that, in the first scenario, social preferences help to reduce tax evasion but (unless they are unrealistically strong) don’t omit the need for tax authorities to proficiently catch under-reporting, and that, in the second scenario, taxpayers will truthfully report even if the probability of being caught if they under-report is relatively low³.

Broadly, the literature on the enforcement of tax compliance can be classified as the studies of punitive and behavioral strategies. An example of the former is the evaluation of the impact of audit rates and rules on compliance (for instance ^{Dubin et al., 1990}; ^{Alm et al., 1993}; ^{Mittone, 2006}; ^{Dubin, 2007}; ^{Kleven et al., 2011}). The latter literature’s branch is relatively recent and it has focused on the study of nudges. For instance, ^{Moulton et al. (2022}), use a field experiment to find that reminders about payment obligations cause a reduction in the probability of missing payments of property tax among older adults who took out a reverse mortgage. Other studies about the effect of nudges on tax compliance include ^{Blumenthal et al. (2001)}; ^{Hallsworth et al. (2017)}; ^{Meiselman (2018)}; ^{John and Blume (2018)}, and ^{Fišar et al. (2022)}. Recent, global views on the extent of tax evasion are given by ^{OECD (2023)} and ^{Alstadsaeter et al. (2024)}. Also, although in contexts not related to Ecuador, from a variety of perspectives, recent contributions have examined tax compliance issues in relationship to filings or sending messages (akin to this paper’s notifications) [^{Carruthers et al., 2022}; ^{Goldin et al., 2022}; ^{Goodman et al., 2023}; ^{Holz et al., 2023}; ^{DeBacker et al. 2024}; ^{Cohen, 2024}; ^{Hauck and Wallossek, 2024}].

More generally, tax compliance is shaped by a range of economic, social, psychological, and institutional factors, as highlighted in the literature reviews by ^{Alm (2019)} and ^{Slemrod (2019)}. Political stability and effective governance positively influence compliance by enhancing enforcement and ensuring the availability of third-party information. Social norms contribute positively when tax compliance is accepted as a societal standard but can have a negative impact in contexts where tax evasion is normalized. Individual participation in government decision-making further encourages compliance, as taxpayers are more likely to comply when they feel involved in the allocation of public resources. Additionally, technological advancements, such as the digitalization of tax systems, play a significant role in improving compliance by increasing transparency, and reducing opportunities for evasion.

This paper contributes to the literature in several ways. First, it is, to the best of our knowledge, the first study that theoretically analyzes the role of unlimited refilings on tax compliance. Second, it shows that, unless a person’s care for the welfare of society as a whole is unrealistically strong, the better decision for taxpayers is to evade taxes if their ability to unlimitedly refill is not curtailed. Third, the paper highlights the importance that policymakers, before enacting policies, carefully consider if policies will correctly incentivize the desired behavior. This is a consideration that has been brought to the forefront in tax collection literature’s several recent papers (^{Yang, 2008}; Alm, 2010; ^{Paula and Scheinkman, 2010}; Gillitzer and Skov, 2013; ^{Carrillo et al., 2017}; ^{Slemrod et al., 2017}; ^{Almunia and Lopez-Rodriguez, 2018}) and -importantly for us to highlight- in several other areas of economics research, including crime economics, ecological economics, labor economics, public sector economics, business economics, and economics of innovation (Becker, 1968; ^{Frey and Oberholzer-Gee, 1997}; ^{Gneezy and Rustichini, 2000}; Fehr and List, 2004; ^{Gneezy and List, 2006}; ^{Mulder et al. 2006}; ^{Bowles, 2008}; Bowles and Hwang, 2008; ^{Ellingsen and Johannesson, 2008}; ^{Georgellis et al. 2011}; ^{Gneezy et al., 2011}; ^{Bowles and Polania-Reyes, 2012}; ^{Kerr et al. 2012}; ^{Madrian, 2014}, ^{Weaver, 2015}; ^{Dai and Wang, 2024}). Finally, it presents a simple policy recommendation to prevent tax evasion that consists of limiting unlimited refilings. Now, while the Ecuadorian tax authority did so since 2013, the earlier situation is a useful phenomenon to study for the very reason that, just as we study history to not repeat our mistakes, a careful theoretical investigation of it could be essential to prevent similar policies from being implemented in the future.

The rest of the paper is organized as follows. Section 2 presents the institutional background around the ITA in Ecuador and the associated empirical findings which motivate our theoretical inquiry. Section 3 presents the formal analysis when unlimited refilings are allowed. Section 4 presents, first, the formal analysis when unlimited refilings are blocked; and, second, the formal analysis in the presence of taxpayers with social preferences. Section 5 concludes.

2. THE INCOME TAX ADVANCE

First, the Ecuadorian tax legislation valid for the fiscal year 2010 and that is relevant for this research is described⁴. The income tax (IT) in Ecuador is similar to the one in the USA. For individuals it is determined using a progressive tax schedule with marginal rates that go from 5 to 35%. For corporations the it is calculated as a flat rate of the tax base that was 25% in 2010.

In addition to the it, Ecuadorian taxpayers have to file the ITA. When taxpayers file their IT _t-1 (in March or April of year t), they also determine the ITA _t . We study corporations and individuals who are required to keep accounting records. For them the ITA _t is calculated as a function of tax records corresponding to the year t-1. Specifically, it is equal to sum of 0.4% of the total assets, 0.4% of the total taxable income, 0.2% of the net worth, and 0.2% of deductible expenses.

It is important to note that the ITA is a minimum income tax. When taxpayers file their IT _t (in the year t+1), the IT _t and the ITA _t are compared, and the greater of the two becomes the relevant income tax (RIT _t ). Moreover, there is an anticipated portion of the ITA _t that is paid in two equal parts in July and September of year t (before the IT _t is determined). This amount is equal to the ITA _t minus taxes withheld (by third parties) in year t-1.

Clearly, there are incentives to evade the ITA. First, a lower ITA _t means a lower anticipated payment and hence more liquidity for the taxpayer in year t. More importantly, a lower ITA _t could potentially imply less paid taxes if the ITA _t ends up being greater than the IT _t . ^{Sánchez (2022)} shows evidence that these incentives are strong, because it documents that in the fiscal year 2010 around 7% of total number of corporations and individually-owned businesses obligated to keep accounting records under-reported the ITA. More details about the ITA in Ecuador are presented in that paper.

3. UNDERSTANDING THE ISSUE

Consider the situation, thus, in which (a) if the taxpayers under-reports their⁶ ITA, then they don’t necessarily get a notification; with p∈[0,1] being the probability that tax authorities send them such notification and 1 - p the probability they don’t; and (b) they are able to freely refile more than once. The latter allows the following two-step (or more) behavior: In a first refiling they could report the true ITA and, just for that, leave the list of evaders, and in a second refiling they could under-report again without any further problem. As shortly shown, in this scenario, to under-report their ITA (assuming it is greater than their IT) is the expected payoff maximizing decision for exclusively selfish taxpayers -i.e., who only care about their individual and monetary payoffs and not about the consequences of evading taxes for society as a whole- even if tax authorities always send them a notification. However, before proceeding, it is important to point out four things. First, because in this scenario taxpayers have greater incentives to under-report the ITA, only the case in which the ITA is greater than the it is examined. However, under reasonable assumptions, the result that under-reporting is the payoff maximizing decision for selfish taxpayers also holds when the ITA is smaller than the IT; see footnote 8. Second, to carry out a general analysis, it is posited that, when truthfully reporting, they still face a probability q∈[0,1] that tax authorities wrongly send them a notification (and a probability 1 - q they don’t). However, it did not happen in Ecuador’s tax control program for the ITA in the period 2010-2012, i.e., in that case q = 0. Third, because it is assumed that the utility that every outcome yields is the same as its payoff and that all payoffs are linear, the models assume taxpayers are risk-neutral. However, see footnote 9. Fourth, it should be warned -and this type of warning applies to any modeling effort- that the inquiry has simplifying assumptions that must not be generalized. Among them, we could mention that, selfish or otherwise, taxpayers maximize expected payoffs and that the tax authority may make mistakes but it is not corrupt.

Logically, under any institutional framework that will be examined, the taxpayer could end up in any of four possible situations: i) report their true ITA and the tax authority sends them a notification, ii) report their true ITA and the tax authority does not send them a notification, iii) under-report their ITA and the tax authority sends them a notification, and iv) under-report their ITA and the tax authority does not send them a notification. Now, in the institutional framework in which (a) and (b) above explained are the case, the payoffs of the selfish taxpayer for each of such situations are as follows. In the first situation, they’ll have to pay their true ITA (denoted by 0 < T) plus the refiling costs of addressing the notification (denoted by 0 < c), thus having a payoff of -(T + c). In the second situation, they’ll only have to pay their true ITA, thus having a payoff of -T. In the third situation, their payoff will be -(T + 2c - u(1 + r)) because of the following. To begin with, they reported 0 < T - u (with 0 < u denoting the amount under-reported). Furthermore, if invested somewhere else, u yields a return of ur (with 0 < r denoting the rate of return). So, if instead of paying u as taxes they could keep it, they would only pay T - u(1 + r). But the fact that unlimited refilings are possible allows them to indeed not pay u(1 + r) in taxes. The only added costs for not paying it would be they have to pay the refiling costs twice (or more times, of course, if they refiled more than twice, but no need to assume so) which are 2c. So, their payoff would be -(T - u(1 + r)+ 2c). In the fourth situation, in addition of being able to not pay u(1 + r) in taxes, they’ll not have to pay refiling costs, thus having a payoff of -(T - u(1 + r)). We assume that 2c < u(1 + r), which is a fairly reasonable assumption given that the under-reporting taxpayer chooses u (with the only restriction, given they can’t report a negative ITA, that u < T).

Therefore, the expected payoffs for reporting the true ITA and for under-reporting it (hereafter, π ^tr and π ^un , respectively) are:

From equations [1] and [2] it follows that:

However, given that 2c < u(1 + r), then π ^tr < π ^un no matter how proficiently (or incompetently) tax authorities carry out their job, that is, for any p([0,1] and q∈[0,1]. Figure 1 left panel graphically shows this result. It has the positively sloped line p(q)q2+u(1+r)2c at which π ^tr = π ^un , which we label p ^unl (for “unlimited refiling”). Above p ^unl it is the case that π ^un < π ^tr , so that reporting the true ITA is the expected payoff maximizing decision. Below p ^unl it is the case that π ^tr < π ^un , so that under-reporting the ITA is the expected payoff maximizing decision. It could be seen that, within the p∈[0,1] and q([0,1] unit square area, it never occurs that π ^un < π ^tr .⁷

Figure 1 The scenarios under selfish preferences 

The intuition of this worrisome result is, actually, straightforward. If accompanied by allowing taxpayers to refile unlimited times, the sending notifications policy does not take away from taxpayers the power to obtain benefits if they under-report; with those benefits being u(1 + r). Under-reporters will only have to pay the costs of refiling twice, 2c. Thus, if the former are greater than the latter, the expected payoff maximizing decision would be to under-report. However, it is reasonable to assume that 2c < u(1 + r), for the very reason that the taxpayer chooses u⁸.

The bottom line of this analysis is that, for selfish taxpayers, allowing unlimited refilings makes the notification policy plainly ineffective⁹.

4. ADDRESSING THE ISSUE

4.2. Social preferences

There is however, a supplementary approach to foster that reporting the true ITA is the expected payoff maximizing decision. It is the presence of taxpayers who, instead of being exclusively selfish, exhibit social preferences. By this, it is meant taxpayers who not only care about their strictly individual and monetary payoffs but they also positively care about the tax revenues raised by society as a whole (see footnote 2). A parsimonious way to model them is assuming that, when reporting their true ITA, in addition of obtaining some monetary payoffs, they also obtain a subjective gain for reporting the truth (so they actually contribute more if it is smaller than ITA). And, before proceeding, let us mention that, although it is beyond the scope of this paper to discuss how these preferences spread among society members, nevertheless, and as ^{Fehr and Schmidt (2006, p. 617)} say, the idea that “people often care about the well-being of others and that this may have important economic consequences” has been pointed out by an array of influential economists, including ^{Smith (1759)}, ^{Becker (1974)}, ^{Arrow (1981)}, ^{Samuelson (1993)} and ^{Sen (1995)}.

To assess the potential role of social preferences on taxpayers’ compliance, it is important to examine their presence both in the institutional framework in which unlimited refilings are possible and in the institutional framework in which it has been blocked. Beginning with the former, consider the case such that, if under-reporting their ITA, the taxpayer payoffs are the same as with selfish preferences (recall: -(T - u(1 + r) + 2c) if receiving the notification and -(T - u(1 + r)) if not); thus, the expected payoff for under-reporting the ITA remains given by equation [2]. However, very differently, given this institutional and preferences framework when the taxpayer reports their true ITA (and regardless of receiving a notification or not) obtains a gain 0 < s. It is the parameter that captures their degree (strength) of social preferences. It is assumed that, on the one hand, u(1 + r) - 2c < s, i.e., social preferences are strong in this degree; but, on the other hand, s < u(1 + r) - c, i.e., social preferences are week in this degree. Therefore, when the taxpayer reports their true ITA and receives a notification, their payoff is -(T + c - s); which is the case because, if they report their true ITA and receive a notification, in addition of having to pay T + c, they would obtain the gain s. If they report their true ITA and don’t receive a notification, their payoff is -(T - s); which is the case because, in addition of having to pay T, they would obtain the gain s.

Therefore, the expected payoff for reporting the true ITA is given by:

πtr=-qT+c-s-1-qT-s=-qc-T+s [6]

From equations [2] and [6] it follows that:

πtr<πun↔p<q2+u(1+r)2c

πun πtr↔q2+u(1+r)2c p [7]

Given that u(1 + r) - 2c < s < u(1 + r) - c, equation [7] implies the following mixed results. On the one hand, social preferences are strong enough so as to achieve that, if tax authorities often send notifications to under-reporters (p → 1) and seldom send notifications to truth-reporters (q → 0), then π ^un < π ^tr . But, on the other hand, social preferences are not strong enough so as to achieve that, if tax authorities always wrongly send notifications to truth-reporters (q = 1), then nevertheless π ^un < π ^tr no matter how often they rightly send notifications to under-reporters (i.e., the fact remains that, if q = 1, then π ^tr < π ^un for any p∈[0,1]).

Figure 2 left panel graphically shows these results. It has the positively sloped line pq=q2+u1+r-s2c at which π ^tr = π ^un , which we label p ^soc,unl (for “social preferences with unlimited refilings”), above/below which π ^un < π ^tr / π ^tr < π ^un and, thus, truth-reporting/under-reporting the ITA is the expected payoff maximizing decision. It could be seen that, within the p∈[0,1] and q∈[0,1] unit square area, there is a subset in which actually π ^un < π ^tr , labeled B. The result that social preferences help to achieve that truth-reporting is expected payoff maximizing but it also requires that the tax authority carries out its job proficiently, could be seen by noting that area B includes the points in which p → 1 and q → 0. The result that π ^un < π ^tr cannot occur if q = 1 (no matter how close to 1 or even equal to 1 p is) could be seen by noting that area B does not include any point on the vertical line q = 1. Furthermore, Figure 2 left panel shows that the vertical intercept of p ^soc,unl is greater than ½, i.e., even if q = 0, to be within B it has to happen that ½ < p.

Figure 2 The scenarios with social preferences 

The intuition of these mixed results is straightforward. Related to the hopeful for tax compliance result, it is the following. If the taxpayers who report their true ITA obtains a gain s that is greater than u(1 + r) - 2c, regardless of whether they wrongly get a notification or not, then it is possible that reporting the true ITA is their expected payoff maximizing decision even if they can still do unlimited refilings. However, for that to be the case, and precisely because unlimited refilings are allowed, it is also needed that tax authorities often send notifications to under-reporters and seldom send notifications to truth-reporters. The intuition of the sobering result is the following. If the gain s the truth-reporter obtains is smaller than u(1 + r) - c, they nevertheless for sure get a notification, and they could also do unlimited refilings, then under-reporting will be their expected payoff maximizing decision.

Moving to the institutional and preferences framework in which, in addition of having taxpayers with social preferences, unlimited refilings have been blocked, in it the expected payoffs for reporting the true ITA and for under-reporting it are given by equations [6] and [4], respectively. From the latter equations it follows that:

πtr<πun↔p<cqc+u(1+r)+u1+r-sc+u(1+r)

πun πtr↔cqc+u(1+r)+u1+r-sc+u(1+r) p [8]

Equation [8] implies the following results. First, for any positive degree of social preferences, π ^un < π ^tr occurs if tax authorities carry out their job fairly well. That is, provided such competent tax authority exists, it is no longer needed that social preferences are strong enough so that u(1 + r) - 2c < s but any 0 < s suffices for truth-reporting to become the expected payoff maximizing decision. Second, and perhaps most importantly, the subset of the tax authority’s performances which achieve that π ^un < π ^tr is the greatest of the four institutional and preferences scenarios we have examined.

Figure 2 right panel graphically shows the latter result. It has the positively sloped line pq=cqc+u(1+r)+u1+r-sc+u(1+r) at which π ^tr = π ^un , which we label p ^soc,one (for “social preferences and one refiling”), above/below which π ^un < π^tr / π ^tr < π ^un and, thus, truth-reporting/under-reporting the ITA is the expected payoff maximizing decision. It could be seen that, within the p∈[0,1] and q∈[0,1] unit square area, there is a subset in which actually π ^un < π ^tr , labeled C. Comparing it with area B of Figure 2 left panel, it could be seen (and formally shown) that C is greater. Furthermore, comparing C with area A of Figure 1 right panel, it could be seen (and formally shown) that the former is greater as well. These comparisons show that, among the four institutional and preferences scenarios that have been examined, that one in which there are social preferences and unlimited refilings are blocked gives rise to the greatest variety of a tax authority’s performances which make that truth-reporting is expected payoff maximizing. Relatedly, Figure 2 also shows that the (right-panel’s) vertical intercept of p ^soc,one is smaller than the (left-panel’s) vertical intercept of p ^soc,unl .¹⁰

The intuition of this hopeful for tax compliance result is straightforward. To begin with, blocking unlimited refilings reduces the expected payoff of under-reporting. In addition, social preferences increase the expected payoff of truth-reporting. Under the combined impact of both effects, reporting the true ITA could become the expected payoff maximizing decision even if the tax authority does not carry out its job that proficiently.

5. CONCLUSIONS

This paper theoretically analyzes the role of unlimited refilings on tax compliance; by which it is meant the ability to refile taxes as many times as desired without going through a formal process that verifies that the refile is legally necessary. It is done in the context of the use of tax notifications to control the evasion of the income tax advance in Ecuador, and the use of unlimited refilings to evade taxes in that country is documented. Building upon this background, a model to study the behavior of taxpayers under different institutional and preferences scenarios is developed.

First, it is assumed that taxpayers exhibit exclusively selfish preferences and unlimited refilings are allowed. The key (and worrisome) finding in this scenario is that, independently of the efficiency of the tax authority to correctly notify the evaders and not to notify compliers, the expected payoff maximizing decision is to evade taxes. Second, it is assumed that unlimited refilings are blocked but taxpayers are still exclusively selfish. The key finding in this scenario is that complying to pay taxes is the payoff maximizing decision as long as the tax authority is strongly efficient notifying evaders. That is, blocking unlimited refilings is a necessary but not sufficient condition to achieve compliance. Third, the scenario in which taxpayers exhibit social preferences and unlimited refilings are allowed is examined. The key finding under this scenario is that there will be compliance, because social preferences foster such a decision, but only if the tax authority delivers the tax notifications efficiently (does not miss under-reporters that often and does not notify truth-reporters). The fourth, last scenario assumes that taxpayers exhibit social preferences and that unlimited refilings are blocked. The key finding in this scenario is that there will be compliance for a broader range of parameters. The need of having an efficient tax authority (i.e., consistently notifies evaders and does not notify compliers) remains but is less stringent the stronger the social preferences are.

Overall, the results imply that blocking unlimited refilings reduces the expected payoff of evasion and therefore reduces the probability of evasion. Furthermore, it is found that social preferences help reducing tax evasion and therefore that the proliferation of society-regarding preferences should be promoted. The findings imply a simple policy recommendation that consist on limiting the option of unlimited refilings in the context of tax evasion control. While the Ecuadorian tax authority blocked it in 2013, its earlier existence (2010-2012) is a useful phenomenon to study for the very reason that empirical and theoretical study of history to avoid repeating mistakes is (or must be) an important research objective.

More generally, the results of this study call for carefully design evasion control policies that are compatible with the incentives of taxpayers. In fact, the call for policymakers to carefully consider, before enacting policies, if the latter will correctly incentivize the desired behavior, is found not only in the tax compliance literature but in several areas of economics research (see the literature referred to in the Introduction section).