CHAPTER 4. Recent Trends and Evidence

Eric Le Borgne, and Katherine Baer
Published Date:
July 2008
  • ShareShare
Show Summary Details

In this section we analyze the recent trends in tax amnesties, and some evidence is provided regarding their revenue and compliance effects. The first section is devoted to a review of the econometric literature on the (mainly revenue) effect of tax amnesties. The second section describes the U.S. states’ experience with amnesties (through a general overview and then two case studies: the tax amnesty programs of Kentucky and Michigan, both in 2002). The experience of the U.S. states is worth investigating because, given the relative homogeneity of these states (at least compared with cross-country studies), it is easier to identify key factors affecting the success of tax amnesty programs. The final section focuses on cross-country experiences, first through a general overview of recent tax amnesty trends, and then through selected case studies (Ireland, Italy, Argentina, Turkey, and the Philippines); these case studies have been selected for the range of amnesty programs they cover.

Econometric Studies

Econometric evaluations of the revenue and compliance effects of tax amnesties are challenging because, in practice, tax amnesties are always combined with enhanced enforcement (i.e., a tax policy measure is combined with a tax administration measure). Statistical identification of each of these effects, rather than the combined effect, is therefore difficult. Bearing this statistical problem in mind, we now review the empirical literature. Most of the econometric studies have been conducted on U.S. states.

Evidence from the U.S. States

Simple statistical evidence shows that even the most successful tax amnesty programs report only a small short-run (one fiscal year) increase in gross tax revenue (Table 2).17 Several factors need to be netted to obtain a proper measure of new tax collection: namely, (1) administrative costs (staff time, advertising budgets, etc.) and (2) “accounts receivable” (this term refers to tax evaders who have already been detected by the tax administration and who have been sent notices of their new tax bill—see Box 2 for details). Accounts receivable should be excluded from the measure of “new money” that results from a tax amnesty because an amnesty merely brings forward the tax payment from these taxpayers, but often at a cost because, absent an amnesty, these detected tax evaders would have had to pay substantial penalties, which are usually forgiven (as part of the tax amnesty incentive package)—see the next section for details.

Table 2.U.S. States’ Tax Amnesties: Summary Statistics on Increased Tax Revenue Collection

(In percent of relevant revenue)1

Standard deviation0.74
First quartile0.09
Third quartile1.05
Source: Calculations based on data from U.S. Congress (1998).

The table covers the period 1983–97 and refers to tax collections in the amnesty as a percentage of the state’s revenue from the taxes covered in the last full fiscal year prior to the amnesty.

Source: Calculations based on data from U.S. Congress (1998).

The table covers the period 1983–97 and refers to tax collections in the amnesty as a percentage of the state’s revenue from the taxes covered in the last full fiscal year prior to the amnesty.

Econometric evidence regarding the (gross) long-run revenue effect of tax amnesty cum increased enforcement strategy is mixed, at best. Joulfaian (1988), studying Massachusetts’ amnesty program, finds that chronic nonfilers as well as filers who underreport tax are not attracted by tax amnesties. Alm, McKee, and Beck (1990), in an experimental study, find that compliance decreases after amnesties (on their own), but rises if amnesties are combined with tighter enforcement. Alm and Beck (1993) find that the 1985 Colorado tax amnesty had no long-run effect on tax revenues (both the level and the growth rate). Christian, Gupta, and Young (2002) find, using data on subsequent filing from the Michigan income tax amnesty, that about two-thirds of “new filers”—that is, taxpayers who either never filed a tax return or did not file for all taxes they were supposed to—and 90 percent of previous filers who filed amended returns under the amnesty program subsequently filed income tax returns. However, the additional revenue raised from these (5,500) taxpayers was marginal (about 0.1 percent of the state’s personal income tax revenues).

Box 2.Accounts Receivable

“Accounts receivable” refers to delinquent taxpayers who have already been detected by the tax administration and who have been sent notices of their new tax bill.

Not surprisingly, tax amnesty programs that allow known delinquent taxpayers to participate in the amnesty are often successful in terms of generating gross revenue. These taxpayers are offered the possibility to settle their taxes owed and detected with the benefit of (1) a waiver of criminal prosecutions; (2) a waiver of civil prosecution; and (3) some financial incentive, usually a reduction in fines and penalties (sometimes a complete waiver), to settle their taxes.

Revenue collected from accounts receivable is the predominant revenue source among U.S. states’ tax amnesty programs.1 Of the $6.6 billion (2004 US$) in total gross amnesty revenue collected by the states during 1980–2004, 92 percent was from amnesty programs that allowed taxpayers with accounts receivable to participate. The average revenue per amnesty is $14 million for programs that do not allow accounts receivable, and it reaches $123 million for those that do. Sixty-two percent of tax amnesty programs allowed taxpayers with accounts receivable to participate in the amnesty.

Taxpayers with accounts receivable and who decide to participate in a tax amnesty reveal (1) that they have the financial means to pay their taxes in full and (2) that they are willing to do so (rather than appeal the tax administration’s tax assessment in court). Therefore, the probability of collecting taxes from these taxpayers is relatively high. For this class of taxpayers, the tax amnesty is likely to

  • entail a loss of revenue in terms of waived penalties; and
  • change the dynamics of collection of the revenue: the amnesty accelerates the collection of some previously assessed taxes that would have been collected in later years.

The net revenue effect from allowing accounts receivable to participate in a tax amnesty is therefore ambiguous (it could easily be negative depending on the generosity of the amnesty program).

Source: IMF staff.1 Notwithstanding this U.S. example, the general issues with accounts receivable are equally valid in developed and developing economies.

Le Borgne (2006), in a recent econometric analysis of the determinants of the occurrence of tax amnesties in U.S. states from 1977 to 1998, extends the study of Dubin, Graetz, and Wilde (1992). These authors find that tax amnesty programs in U.S. states are more likely to be driven by the potential yield (high taxes) that they can produce rather than by the need to increase revenue in light of fiscal duress. As Dubin, Graetz, and Wilde note, Le Borgne finds supportive evidence that states with high tax levels are (slightly) more likely to initiate a tax amnesty. However, in contrast with Dubin, Graetz, and Wilde, Le Borgne finds that, even controlling for the revenue yield effect (and a state’s growth rate), tax amnesties are more likely to be declared when a state is experiencing a growing budget deficit. Tax amnesties are therefore viewed as a revenue-raising source. Le Borgne also finds that the timing of tax amnesties is affected by political considerations. In keeping with the well-known aversion to increase taxes ahead of elections (Besley and Case, 2003), state governors are far less likely to declare a tax amnesty during a gubernatorial election year. Because of the peculiarity of tax amnesty programs as a revenue source—that is, raising revenue from taxes owed in the past—these programs could conceivably be even more costly politically because law-abiding taxpayers/voters might consider a tax amnesty as “unfair” to them. The evidence reported in Le Borgne is consistent with this view: there is a significant correlation between incumbent governors losing their reelection bid and declaring a tax amnesty during their reelection year.

Recent Econometric Evidence from Around the World

Torgler and Schaltegger (2005), using a study of tax amnesty programs in Switzerland and Costa Rica, analyze the relationship between tax compliance and citizens’ ability to vote in favor (or not) of introducing a tax amnesty program. They find that tax compliance only increases after voting. Voting offers citizens an opportunity to discuss the merits and purpose of the amnesty program; the authors conjecture that this may induce a kind of civic duty, as taxpayers become aware of the importance of contributing to the provision of public goods.

López-Laborda and Rodrigo (2003) investigate the long-term impact on Spanish individual income tax (IRPF) compliance of the amnesty measures granted in 1991 within the framework of the 1988–91 income tax reform program. Using advanced time series techniques (testing for the endogenous estimation of structural breaks) on monthly data from 1979 to 1998, they find that the amnesty had no effect on tax collection in either the short or the long term. However, they do find evidence of a permanent positive impact caused by the legislative and administrative measures linked to the IRPF reform process that started in 1988.

Das-Gupta and Mookherjee (2000) provide empirical estimates of the revenue effects of various tax amnesty programs offered in India from 1965 to 1991. From their empirical analysis, they find that (1) the early tax amnesty programs (up to 1980) were arguably unanticipated by taxpayers; however, apart from the 1975 program, they all produced negative revenue gains, once the indirect effects (compliance, filing behavior, waiving of fines, and penalties, etc.) were taken into account; (2) post-1980 programs had all been widely anticipated and therefore resulted in a negative overall revenue effect; and (3) the revenue impact of tax amnesty programs declines over time.

Trends and Case Studies

Global Trends

Amnesties have been used repeatedly over time and across countries, regardless of their degree of economic development. Developed countries such as Australia, Austria (1982, 1993), Belgium (1984), Finland (1982, 1984), France (1982, 1986), Greece, Ireland (1988, 1993), Italy (1982, 1984, 2002), New Zealand (1988), Portugal (1981, 1982, 1986, 1988), Spain (1977), and Switzerland have all introduced amnesties at some point. Developing countries that have also enacted tax amnesties (often repeatedly) include Argentina (1987, 1995), Bolivia, Chile, Colombia (1987), Ecuador, India (several years), Panama (1974), Peru, Mexico, the Philippines, the Russian Federation (1993, 1996, 1997), and Turkey.18 In the United States, although states did not offer a tax amnesty until 1982, they have since offered some 78 programs (until 2004), with almost all states (42 out of 50) offering a tax amnesty of some sort. Most of these U.S. states have introduced several amnesties during that period (e.g., four for Louisiana and three for Arizona, Connecticut, Florida, Massachusetts, Missouri, New Jersey, and New York). As shown in Figure 1, amnesties tend to be clustered over time. Figure 2 reveals that revenue collection from amnesty programs also tends to have a similar clustering.

Figure 1.U.S. States: Number of Tax Amnesties Per Year, 1982–2004

Source: IMF staff calculations based on data from the Federation of Tax Administrators (FTA).

Figure 2.U.S. States: Tax Amnesty Gross Revenue Per Year, 1982–2004

(In 2004 U.S. dollars)

Source: IMF staff calculations based on data from the FTA.

International evidence regarding the effectiveness of tax amnesties is mixed. “Successful” tax amnesties—at least based on the gross amount of revenue raised—include the Argentine general amnesty of 1995 (whose gross yield was around $3.9 billion), an amnesty in India in 1997 ($2.5 billion) according to Time Asia (1998), the Irish amnesty of 1988 ($700 million), and Italy’s 2002 amnesty on capital repatriation. In the United States, the gross revenue collected from the 78 amnesties during the period 1980–2004 totaled $6.6 billion (in 2004 U.S. dollars).

Repeated tax amnesties tend to be less successful in raising gross tax revenues. Recently, Ireland offered a series of tax amnesties (1988, 1993, and 1999); Italy adopted 27 amnesties over a roughly 20-year period; the Philippines had about 15 amnesties over 11 to 12 years in the 1980s and 1990s, and compliance declined over that time. India has a long history of amnesty.

We now turn to selected case studies, starting with two tax amnesty programs offered in 2002 by two U.S. states, Kentucky and Michigan. Both of these programs have been described as a success. The program designers took into account the experience these states had with previous tax amnesty programs. Next, we analyze country case studies. We start with Ireland’s experiences with tax amnesties, which include some very successful programs and one less successful one. We then analyze Italy’s 2001 tax amnesty (Scudo Fiscale) on offshore investment tax evasion. This amnesty program has received a lot of attention, mainly from other European countries, for it is thought of as having been quite successful (in part because of the very large offshore capital that was reported as repatriated). However, against its stated goals of increasing the tax base and fostering domestic investment and growth, the impact of the amnesty is limited, at best. We then turn to Argentina’s various amnesty programs from 1995 to 2004; because of the high frequency of occurrence of these programs, the overall impact of the various programs offered is negative. We then investigate Turkey’s 2003 amnesty programs aimed at reducing the stock of tax and social contribution arrears. The results of these programs have not been positive. Finally, we study the Philippines’ multiple experiences with tax amnesty programs from 1972 to 1987. As in the Argentine case, the high frequency of amnesty programs has probably reduced overall tax compliance and contributed to the continued decline in the tax-to-GDP ratio.



Faced with an economic downturn and a projected reduction (in nominal terms) in General Fund (GF) fiscal revenue for fiscal year (FY) 2002, the governor of the U.S. state of Kentucky decided to introduce a tax amnesty in 2002. The aims of the program were twofold: first, to raise revenue in the short term so as to loosen the fiscal constraint (i.e., Kentucky has a strict balanced budget requirement for its GF); second, to broaden the tax base and improve future compliance.

The key features of the tax amnesty program were20 (1) generous incentives to participate, (2) a significant increase in the costs of tax evasion, and (3) a widespread public information campaign. Among the incentives were a waiving of all interest and collection fees on taxes owed as well as criminal prosecution and all civil penalties. The added tax-evasion costs included (1) collection fees increasing from 20 to 25 percent, (2) a new 25 percent assessment fee on all additional liabilities assessed for any amnesty-eligible period, (3) a new 50 percent failure-to-file fee, (4) the possibility of publishing the list of delinquent taxpayers, and (5) corporate officers and members of limited liability and limited partnership companies being individually liable for payment of some evaded taxes. The operational cost of the tax amnesty (mainly advertising) reached $2.8 million (lower than the allocated $3.1 million).

The 2002 tax amnesty was Kentucky’s second; the first one dated back to 1988. That first amnesty provided for the waiver of criminal prosecution, all civil penalties, and one-half of the interest due (compared with 100 percent in 2002). A total of 18,761 taxpayers participated in that first amnesty program and paid a total of $61.1 million—representing 1.8 percent of total GF revenue (which reached $3.344 billion in 1988).

Effects and Assessment

The 2002 amnesty program attracted a large number of participants and raised significant revenue. A total of 23,592 taxpayers’ applications were received in 2002 (52 percent from individuals, the rest from businesses).21 Of these 23,592 taxpayers, only 346 applicants (i.e., less than 1.5 percent of the total) were new filers. The vast majority of amnesty revenue came from businesses (86 percent, or $103.7 million). Total revenue collection reached $123.4 million.

Despite the appearance of success, the amnesty program was disappointing as assessed against its objectives. Total collection did reach $123.4 million, which is a significant amount, especially in times of fiscal duress. However, this headline number is deceptive (as a judge of success) for the following reasons:

  • Hardly any new filers participated in the amnesty. A key goal of the tax amnesty was to increase tax compliance and broaden the tax base. On the basis of this key criterion the amnesty program seemed to have had, at best, a limited impact: both the number of new filers (346) and the amount they paid ($2.6 million: 2.1 percent of total amnesty revenue or 0.03 percent of total GF revenue) are negligible.22
  • A significant share (37 percent) of the $123.4 million came from accounts receivable. Most of these revenues would have been collected over time; the amnesty just brought these revenues forward at the expense of subsequent fiscal years (especially FY 2003).
  • The amnesty is likely to have decreased the compliance level of taxpayers who were current with their liabilities. Indeed, according to information gathered from the 2002 tax amnesty program, the 1988 tax amnesty led to a decrease in compliance in 1989. Given the expected lack-of-recall problem (i.e., people forget activities that took place long ago, which in this context includes taxes owed in the distant past), one would expect people—for a given degree of compliance—to have a better recollection of their recent tax-evasion activities. Figure 3 shows that the lack-of-recall problem is indeed acute (or that taxpayers estimate that the probability of detection declines rapidly as years pass), except for FY 1989 (1 percent of total collection, versus 0.5 percent for both 1990 and 1991).23
  • The 2002 amnesty yielded slightly less revenue than did the 1988 program despite more generous participation incentives and increased tax-evasion penalties. Indeed, the $123.4 million in tax amnesty payments amounted to 1.7 percent of the state’s GF revenue ($7.4 billion in 2002) compared with the 1.8 percent collected during the 1988 tax amnesty.

Figure 3.Kentucky: Time Allocation of the Revenue Collected Through the 2002 Tax Amnesty by Year of Tax Evasion, 1989–2001

(In percent)

Source: Kentucky Revenue Cabinet.


Kentucky’s 2002 tax amnesty program did not meet one of its two goals (broadening the tax base and improving future compliance). However, it did manage to raise revenue in the short term (tax amnesty payments amounted to 1.7 percent of the state’s 2002 GF revenue), which provided the state some fiscal breathing space in a difficult economic environment and, at the margin, therefore helped the state either not to increase taxes or not to cut expenditures, given its balanced-budget requirement. The (significant) downside of this short-term revenue gain is its cost: a large share of these revenues stemmed from identified tax evaders who would have eventually paid most of their tax liabilities; compliance, going forward, might also have been negatively affected, as data from Kentucky’s previous tax amnesty indicate.



In 2002, the U.S. state of Michigan introduced a tax amnesty, motivated by a nominal decline in the state’s revenue (two years in a row) and also a desire to broaden the tax base. After a peak in revenue in 2000, GF revenue declined, in nominal terms, by 2 percent in 2001 to $9.8 billion and, from 2001 to 2002, by another 5.5 percent (to $9.3 billion). This left the state’s public finances in a difficult situation because, by law, the governor has to submit a balanced budget, and the legislature is required to pass a balanced budget.

The key features of the 2002 program were25 (1) moderate participation incentives (although it provided for a waiver of penalties and prosecutions for program participants—provided filing and payments were received before July 1, 2002—interest due was not waived); (2) a moderate increase in the future costs of tax evasion (after the amnesty period ended, failing to file a return incurred an additional 25 percent penalty); and (3) a limited public information campaign.

As in Kentucky’s case, the Michigan 2002 tax amnesty was the state’s second; the first one dated back to 1986. Michigan’s first tax amnesty provided for the waiver of criminal prosecution, all civil penalties, and interest due. A substantial public relations campaign (“Get to us before we get to you”) took place. All types of state taxes were covered by the 1986 amnesty; accounts receivable were eligible. In gross revenue terms, the 1986 amnesty was considered a success: it generated $109.8 million, with $44.6 million coming from new taxpayers and the remaining $65.2 million coming from taxpayers with assessments. In relative terms, the 1986 amnesty brought in 1.7 percent of total GF revenue, while amnesty revenue from new taxpayers amounted to 0.7 percent of GF revenue.

Effects and Assessment

A total of 22,220 taxpayers participated in the 2002 amnesty,26 of which 4,225 were new taxpayers (they filed a total of 13,854 tax returns) and 15,995 were taxpayers with existing assessments (they filed 53,001 returns). Most taxpayers filed for the individual income tax (10,095 taxpayers, of whom 2,482 were new filers), with income tax withholding (3,452 taxpayers) and sales tax (3,067 taxpayers) being the other major taxes for which taxpayers decided to take advantage of the amnesty.

The amnesty was successful in terms of gross revenue generated from new filers (compared with expectations). Indeed, actual revenue collected from new filers reached $31.7 million, significantly exceeding the expected target of $24.3 million (out of this, $1.5 million was set aside for program expenses). Also worth noting was the high share of revenue collected from new filers out of total tax amnesty collection (39 percent). Further analysis reveals that, out of a total gross revenue collection of $81.9 million, more than a quarter ($20.5 million) came from interest paid on the delinquent taxes (of which $6.9 million came from new filers), and that business taxpayers contributed to 77 percent of the tax amnesty payments.

The apparent success with gross collection from new filers, however, needs to be considered in light of the costs of the program, which include

  • Static costs: the amnesty program waived $22.9 million in fines and penalties on accounts receivable. Thus, the static cost of collecting the $31.7 million in new revenue was both the $1.5 million in program expenses and the $22.9 million in waived penalties that potentially could have been collected at some point in the future.
  • Dynamic costs: intertemporal shifts in tax revenue. The dynamic effects of the tax amnesty are complex and difficult to assess. Some of the effects that can be estimated are the following:
    • Anticipated tax amnesty. Such an expectation could lead to a drop in current revenues, as taxpayers estimate the probability of being detected in the short time until the expected amnesty as sufficiently small, so tax evasion becomes the utility-maximizing choice (in essence, if interest is waived or reduced, it becomes a free or reduced-interest-rate loan from the state to tax evaders). In Michigan’s case, this problem was mitigated by announcing simultaneously the forthcoming tax amnesty (about a year before its introduction) and the fact that the amnesty would include only taxes due up to the day of the announcement of the amnesty.
    • Accounts receivable bringing revenue forward. An amnesty is the opportunity for some taxpayers to resolve their dispute with the tax administration by avoiding penalties; the amnesty therefore accelerates the collection of some previously assessed taxes that would have been collected in later years (mainly FY 2003). The Office of Revenue and Tax Analysis of the Michigan Department of Treasury estimated that, out of the $50.2 million collected from identified delinquent taxpayers, $18.4 million was brought forward from FY 2003 revenue, thereby reducing FY 2003 revenue by that amount.
    • Compliance effect (a complex, yet crucial, dynamic effect). Tax amnesties can either increase compliance by enabling current tax evaders to regularize their situation or decrease it by increasing expectations of future tax amnesties and by discouraging current law-abiding citizens (who observe that previous tax evaders were not detected). In Michigan, tax amnesties seem to have had, at best, a limited effect on compliance. According to indicative data from the 2002 tax amnesty, the 1986 amnesty does not seem to have significantly increased noncompliance because most of the 2002 amnesty payments related to recent tax years.27 This situation is, however, to be expected owing to the well-known recall problem. However, given that collection from new filers, excluding interest, amounted to $24.8 million (which corresponds to several years of taxes owed), the net effect on FY 2003 is likely to be either small or negative (taking into account the $18.4 million in accounts receivable brought forward to FY 2002 and some increase in noncompliance by currently compliant taxpayers who find the current tax amnesty unjust to them). For outer fiscal years, the net effect could be positive, albeit small, because of the phasing out of the revenue brought forward from current accounts receivable.


Overall, the 2002 tax amnesty had only half the success of the 1986 tax amnesty. Although the 2002 tax amnesty raised $81.9 million of gross revenue, this amount represents less than 0.9 percent of the state of Michigan’s 2002 GF revenue, and only about half the 1.7 percent of GF revenue collected during the previous—and first—tax amnesty of 1986. Similarly, amnesty revenue from new taxpayers amounted to 0.3 percent of GF revenue, less than half the 0.7 percent collected during the 1986 tax amnesty.



Ireland’s public sector finances were in a precarious situation toward the end of the 1980s (e.g., the public debt-to-GDP ratio peaked at 147 in 1988). The tax administration (the Irish Revenue, IR) was also widely seen as inefficient. In 1998, as part of a wide-ranging public sector consolidation and reform effort, the IR decided to reform its “business processes” fundamentally (e.g., it introduced self-assessment of taxpayers’ liabilities and undertook a large investment in information technology). To assist this reform effort and start anew in its relationship with taxpayers, three major types of tax amnesty/voluntary disclosure programs were introduced: (1) the 1988 interest rate amnesty, whose goal was to clear the large stock of tax arrears; (2) the 1993 “underreporting” amnesty;28 and (3) a series of voluntary disclosure schemes (1999 onward).

The 1988 Amnesty

The 1988 amnesty focused on clearing up a large stock of arrears (which stood at close to 40 percent of annual tax revenues at that time; much of this debt was estimated, because proper tax returns were not available). The amnesty offered a waiver of interest and penalties if tax arrears were paid by September 30, 1988. In parallel with offering this tax amnesty, the IR undertook a major revamping of its tax arrears collection procedures (along with other, more general reforms of its “business processes”), with a view to preventing a renewed growth of the stock of arrears in the future.

The 1993 Amnesty

The 1993 amnesty allowed tax liabilities on unreported income to be settled with flat-rate payment of 15 percent of the income; interest and penalties were waived in their entirety. An interesting feature of this amnesty was anonymity: tax inspectors were not allowed to know who applied for the program. If a tax inspector audited an amnesty claimant for the years covered by the amnesty, production of a summary amnesty certificate meant that those years were off limits to the tax administration unless it had good reason to believe that the amnesty disclosure had not been complete.

The 1993 amnesty was accompanied by increased penalties (including an eight-year prison sentence in cases where the amnesty was not properly used and the tax underpaid was more than IR£100,000). However, the anonymity clause made evaluating the veracity of declarations made in connection with the amnesty, or indeed getting the evidence to show that the amnesty had not been used at all, very difficult.

The Voluntary Disclosure Schemes (1999 onward)

These schemes were not the most commonly used tax amnesties in that they did not include any financial incentives to program participants: all tax and interest generally had to be paid in full. However, they provided for legal forgiveness (a waiver from civil and criminal prosecutions). The IR was able to offer the voluntary disclosure schemes without financial incentives because of strong new administrative powers it obtained in 1999 (providing the IR with access to previously secret bank accounts). The IR notified (in advance) target groups that it intended to use these new powers, which encouraged voluntary disclosure before a tax investigation was started. These powers proved very fruitful (yielding more than €1 billion); see Appendix I for more details on these schemes.

Effects and Assessment

  • The 1988 amnesty successfully met its goal of clearing the large stock of tax arrears and, in the process, collected significant revenue. The stock of arrears went down from 40 percent of annual revenue to an annual level of approximately 2.5 percent. Also, gross revenue collected reached yielded IR£517 million (1.9 percent of GDP) through some 350,000 payments.29 The IR had forecasted a yield of just IR£50 million.
  • The 1993 amnesty was also relatively successful in terms of gross revenue collection (IR£185 million). However, large numbers of people did not come forward to pay even the 15 percent. This was evident from the number of taxpayers who subsequently came forward under the voluntary disclosure schemes that started in 1999. One possible reason is that the tax administration was perceived as having limited powers and, at the time, low credibility in terms of follow-through: the amnesty program did not raise the cost of tax evasion sufficiently compared with the benefits to affect overall tax compliance fundamentally.
  • Powers and credibility came in 1999, after a major public inquiry into the tax administration and the financial institutions. After this, the banks became much more cooperative and the IR was determined to use its new powers to full effect to restore public confidence in the tax administration (see Appendix I for details).
  • The capacity to “leverage” these powers and credibility, and to take full advantage of the banks’ new-found cooperation, became evident in the program for investigating bogus nonresident bank accounts. The tax administration’s success in this allowed it to develop a new model, which facilitated the efficient cleanup of “legacy” evasion involving offshore subsidiaries of Irish banks and safe-haven insurance products.
  • This new voluntary disclosure model is fairly widely accepted because it involves considerable pain for the tax evader; in many cases the impact of full interest and penalties (penalties can be mitigated for post-1993-amnesty years only) means that voluntary settlement amounts are typically three or four times the evaded tax.
  • The cooperation of the financial institutions has also been very important: if a bank or insurance company can be persuaded to write directly to its customers indicating the benefits of voluntary disclosure, this is the strongest possible signal that the bank will provide the tax administration with the relevant tax-related information when the formal investigation starts.
  • Other enforcement powers have also been effective. For example, the “name and shame” provisions have been very important: every quarter the tax administration publishes long lists of names and settlement amounts identifying those who did not come forward voluntarily. This regularly reinforces the effectiveness of the tax administration’s new powers and its resolve to follow through.


Ireland’s tax amnesty programs, especially the voluntary disclosure schemes, have had a positive impact on reducing noncompliance and on driving long-term revenue upward.30 Key to this success, however, was not the tax amnesty programs per se, but the much improved enforcement capacity that was developed prior to the launch of the programs. More effective enforcement programs led to a significant and structural shift in the cost-benefit analysis of tax evasion in Ireland: the improved capacity of the tax administration to detect evaders and follow through, the increased penalties for tax evasion, and strong administrative and legal enforcement powers all led to a sharp increase in the cost of tax evasion (higher expected probability of detection, higher penalties and fines if caught, and higher likelihood of being forced to pay if detected). At the same time, through a series of large cuts in the top marginal tax rates, the benefits of tax evasion have also been reduced. The combined effort to significantly increase the cost and lower the benefits of tax evasion has resulted in a large improvement in tax compliance.

Italy: The Scudo Fiscale31


To herald a “new regime” (the launch of the euro and the introduction of a number of tax reforms), in 2001 Italy introduced the Scudo Fiscale, a tax amnesty aimed at repatriating unreported offshore investments.32 The objectives of this amnesty were threefold: (1) to broaden the tax base (and therefore tax revenue), (2) to boost the supply of financial resources available for domestic investment by facilitating the return of assets previously held abroad, and (3) to boost economic growth (as a result of increased domestic investment). Prior to the introduction of the euro, many Italians were believed to be keeping some of their assets in offshore financial centers out of concern about inflation, weak currency, political instability, and high taxation on financial income. The Scudo Fiscale was offered as an opportunity for tax evaders to regularize their tax situation at the time of a regime change (the introduction of the euro removed both the currency and inflation concerns, and a series of tax reforms had reduced the effective tax rate on financial income).

Participation incentives were very generous and the program targeted a wide audience (all individual taxpayers resident in the country were eligible). Tax amnesty was granted provided tax evaders paid a light one-off penalty (2.5 percent of the funds declared to be held abroad);33 payment of this 2.5 percent penalty excused participants from all related tax obligations, social security contributions, interest, and penalties, as well as administrative and criminal sanctions. Participation in the amnesty program also protected the privacy of participants, granting them full anonymity. Participants had to repatriate foreign funds through domestic financial institutions (which would receive and pay to the tax authorities the one-off penalty), which were under the obligation of not revealing to the tax authorities the taxpayer’s identity, the amount repatriated, or the associated income. Finally, taxpayers had the option of repatriating capital (repatriation) or simply declaring capital held abroad without repatriation (regularization). In the first case, those who maintained their assets with a financial institution were excused from future income tax declarations and thus were able to maintain their anonymity over time. In the event of an audit on past periods, taxpayers could avoid the audit up to the amount of the regularized or repatriated capital.

The costs associated with continued tax evasion through offshore investments were increased substantially. The penalty for undisclosed foreign assets increased from a fixed penalty of €520 to a penalty rate of between 5 and 25 percent of the value of the undisclosed assets. Moreover, the government committed to make offshore tax fraud detection a priority.

Effects and Assessment

Overall, the gains achieved by the Scudo Fiscale, with one exception, fell short of expectations:

  • Although the Scudo Fiscale managed to broaden the tax base in the short term, the revenue gain was limited. The authorities expected €40 billion to be regularized, with an associated revenue gain of 0.9 percent of GDP (Banca d’Italia, 2002). The outturn exceeded projections as far as assets regularized were concerned (€60 billion or 4.5 percent of GDP), but the short-term impact on revenue was only about 0.1 percent of GDP, and the projected structural increase in revenue was less than 0.03 percentage points of GDP a year.34
  • Based on balance of payments data, the effect of the Scudo Fiscale on net capital repatriation was limited. Though €60 billion was declared to the tax authorities as repatriated, the balance of payments registered limited positive net capital inflow: in 2002, residents’ assets abroad decreased by only €11.6 billion, compared with an increase of €22.3 billion in 2001, with reductions being mainly concentrated in the first semester of the year when the amnesty was effective (see Banca d’Italia, 2003). Possible explanations for this contradiction include that (1) only two-thirds of declared capital was actually repatriated (3 percent of GDP); (2) even if assets were effectively repatriated, taxpayers had no obligation to keep them in Italy so that they could (immediately) reexport repatriated assets, without losing the benefits of the tax amnesty; (3) even if taxpayers did not reexport capital, financial institutions (where assets were deposited) could have exported them as part of their own portfolio strategy; and (4) taxpayers could have anticipated the tax amnesty, taking capital abroad in order to regularize their evaded assets.
  • Also, the Scudo Fiscale does not appear to have had any perceptible impact on investment or growth.35 Several reasons account for this: (1) the capital inflow was limited compared with the size of the economy; (2) the amnesty incentives induced taxpayers to maintain their repatriated capital in the form of financial rather than physical investment (this guaranteed future anonymity with respect to the tax authorities); and (3) the expectation that repatriated capital could have been invested in the productive process at home relied on the assumption that taxpayers did not have alternative sources of investment for their assets (Cottarelli, 2002).

In contrast to the limited benefits achieved, the revenue costs of the Scudo Fiscale were large and opportunities to improve revenue administration were missed:

  • The tax discount was very large. The amnesty provided two types of tax discounts: (1) a discount on the income derived from offshore capital and (2) a discount on the capital itself. Regarding the first discount, tax amnesty participants had to pay, in lieu of taxes and penalties, 2.5 percent of declared offshore capital. This was equivalent to paying tax on the interest matured on that capital for a period of about four years (at a gross return of 5 percent with a final withholding tax of 12.5 percent). But the taxpayer might have held the capital abroad for a much longer period of time, thus receiving a tax discount. Regarding the second discount, the tax amnesty did not include any tax on repatriated capital. Therefore, in the likely event that the repatriated capital was the result of previous income tax evasion, the amnesty also granted full exemption from all taxes potentially levied on that capital. Indeed, in the event of an audit, the taxpayer participating in the amnesty could avoid tax disputes related to any tax up to the amount of the repatriated capital. These two discounts gave rise to a large overall tax discount.36
  • The Scudo Fiscale did not improve the ability of the tax authorities to detect tax evasion. Because the amnesty program granted full anonymity to participants repatriating offshore assets, tax authorities could not access information that could be used to improve audits and detect future evaders.
  • Going forward, incentives for tax amnesty participants to remain compliant were limited. Taxpayers were granted full anonymity regarding their repatriating offshore assets, and could maintain this status by holding repatriated assets in the form of financial investments with a financial intermediary.
  • Future tax compliance is unlikely to improve. The tax amnesty regularized previous evasion cases but did not substantially modify the evasion incentives built into the tax system. First, the structure of the tax system did not change vis-à-vis other countries, leaving unchanged the reasons to evade. Second, despite the increase in evasion penalties, the short-term probability of detecting tax evasion did not change significantly, leaving the cost of evasion rather low compared with the potential financial benefits of tax evasion. This, together with the large tax discount and the fact that the authorities have granted numerous other tax amnesties, may well make evasion more attractive and have a negative long-term effect on revenue collection.37


The Scudo Fiscale generated limited benefits at a very high cost. It was introduced at a propitious time (a regime change) and did provide an incentive for the return (at least temporarily) of large amounts of undeclared offshore investments. However, it granted very large tax discounts and thus had limited effects on revenue collection. Indeed, the long-term revenue effect could well be negative because (1) the Scudo Fiscale was not combined with an improvement of the revenue administration; (2) it guaranteed taxpayers full anonymity, thereby limiting the ability of the tax authorities to collect information and detect future evasion; and (3) it might have reinforced expectations of a future tax amnesty (a second window was indeed offered in 2003).38



Argentina had approximately a dozen tax amnesties between the mid-1950s and the early 1990s. It granted amnesties roughly every other year over the 1977–87 period. Most of these amnesties coincided with changes in government. More recently, seven tax amnesty programs were launched from 1995 to 2004.

Argentina has experimented with various types of tax amnesties. Although all these programs provided for a reduction of penalties and interest, the other participation incentives varied significantly. The key programs and their key features are the following:

  • The blanqueos provided for a reduction of the basic tax liability. From 1970 to 1992, congress decreed seven blanqueos.
  • The moratorias are time-limited amnesties, implying reduction of penalties and interest, and can be implemented as “spontaneous filing” (voluntary filing) and as “payment facilities.” Payment facilities were introduced in 1990 to include obligations that had not been included in the spontaneous filing regime. The main difference between the two types of amnesty is that the spontaneous filing regime does not benefit taxpayers undergoing tax audits. The mechanism is simple: the legislation defines a cut-off date—to consolidate the tax obligations—and a deadline for filing under the special conditions of the amnesty. The legislation specifies the maximum number of payments, the interest rate, and the minimum amount of each payment.
  • In 1991, an “audit amnesty” was introduced to “protect” taxpayers against audits and to encourage compliance with current tax obligations. The program limited the tax administration’s capacity to audit taxpayers for a certain period of time (for tax periods prior to 1991), as long as they were complying with their current obligations.
  • In 1997, an amnesty program contained (1) a regime of voluntary filing of tax returns to encourage taxpayers to declare taxes and social security contribution liabilities that were more than one year old, (2) a reduction in the interest rate for late payment of up to 50 percent, (3) a waiver of financial penalties, and (4) a payment installment program for the remaining tax liabilities.
  • In 2000, an amnesty program provided for (1) a regime for consolidating tax debts; (2) the elimination of the regime of voluntary filing of returns; and (3) a complete waiver of financial penalties and sanctions, and a partial waiver of interest penalties for late payment.
  • In November 2001, a tax amnesty provided participants who paid 1 percent of their tax arrears by December 2001 eligibility for partial forgiveness of interest due and full cancellation of fines and penalties. Participants could also reschedule payment of their remaining arrears under payment facility plans. By February 2002, no debtors had met the conditions to be entitled to the tax amnesty. With tax revenue continuing to fall, the tax administration decided to extend the amnesty via a so-called “mini-amnesty” for unpaid taxes that accumulated between October 2001 and January 2002.
  • In 2003, a tax amnesty program provided for (1) a 50 percent reduction in overdue interest payments, which were also capped at 30 percent of the principal value of the tax obligations, and (2) a waiver of all financial penalties. It also capped the interest rate on the repayment of the principal amount of tax owed (which could be paid over five years) at 6 percent annually, compared with the prevailing monthly rate of 3 percent.

Effects and Assessment

Though it is difficult to assess these programs’ overall revenue impact, effectiveness, and costs (especially in the medium to long term, as amnesty programs were so frequent and other factors affected tax revenue collection), some major observations can be made:

  • There are indications that repeated recourse to tax amnesties has undermined voluntary compliance and has reduced the amnesties’ effectiveness over time. For example, starting in 1990, taxpayers who had unpaid installments from past amnesties could consolidate these debts during subsequent amnesties; on several occasions, additional decrees were issued to allow post-cut-off-date liabilities to be included in an ongoing amnesty (Cuevas, 2000). The ability to roll over tax debts has doubtless had a negative effect on taxpayer compliance.
  • Were it not for the introduction of new taxes, the tax-to-GDP ratio would have fallen by an estimated 2½ points between 1996 and 2000. The declining trend in the tax-to-GDP ratio during this period reflected a fall in economic activity, but is also likely to reflect a decline in compliance rates.39
  • Gross payments generated from recent tax amnesty programs have been modest, hovering around 2 to 4 percent of total tax collection from 1995 to 2003 (Table 3). Moreover, and as should be expected, since 1995 revenue from amnesties has been declining as a percent of GDP to nearly negligible levels, which suggests that their effectiveness as a revenue-raising instrument has been declining. Also, given the frequency with which they have been used (amnesties are practically a standard administrative feature of the tax collection system), there appears to be a kind of “tax amnesty fatigue” among taxpayers and a strong belief that the next amnesty is coming soon so that tax noncompliance can always be rectified promptly, if needed.
  • The fact that, at the height of the country’s economic crisis in late 2001 and early 2002, no taxpayers had complied with the conditions of the (latest) tax amnesty suggests that other administrative arrangements (e.g., a well-structured payment installment program) may have been more effective in motivating taxpayers to settle their overdue tax arrears.
  • The tax amnesties that were offered to encourage previously unregistered taxpayers to register to pay taxes (2000, 2001, and 2002 amnesties) failed to achieve this objective. A 2005 study by the Argentine tax agency found that only 551 new taxpayers registered to pay taxes as a result of the amnesty programs offered since 1995.40
  • Two major reasons for introducing so many tax amnesty programs are weak tax administration procedures41 and an inadequate penalty structure.42
  • Managing these amnesty programs has implied costs for the Argentine tax administration, including lost time and resources, because of the interruption of regular administrative procedures (in particular, audit and enforcement activity). However, the tax department has not measured these costs or taken them into account when considering new tax amnesty programs.
  • Repeated tax amnesty programs in Argentina have adversely affected the equity and fairness of the tax system. Taxpayers who comply with their obligations on a regular basis see potential benefits from noncompliance in the form of either reduced tax payments via reliance on tax amnesties or long-term noncompliance in light of the revealed lack of strong administrative capacity. Noncompliers face fewer incentives to change their behavior given the prospect of repeated tax amnesties and the low likelihood that they will face enforcement actions.
Table 3.Argentina: Payments Made Through Tax Amnesty Programs, 1995–2003(In percent)
Amnesty payments/total tax collection1.
Amnesty payments/GDP0.2800.1600.0500.0100.0030.0020.0010.0300.005
Sources: National Agency of Public Revenue of Argentina; and IMF staff estimates.
Sources: National Agency of Public Revenue of Argentina; and IMF staff estimates.


Argentina’s tax amnesty programs have not been successful, as revealed by the need to frequently introduce new programs aimed at boosting tax compliance and by the results of some of the tax amnesties, which have recently been analyzed by the Argentine tax department. Most amnesties were introduced independently of significant tax policy changes or of an overall program to strengthen the tax administration’s detection and enforcement capacity. Thus, even though it is difficult to measure the effects of the repeated amnesties on overall tax compliance because of the effect of other factors (tax rate increases, new taxes, etc.), these programs do not appear to have contributed in any evident way to improve tax compliance and, on the contrary, are likely to have negatively affected taxpayers’ perception of the fairness and effectiveness of the tax system.



Successive governments have relied on tax amnesties to reduce arrears, both of tax payments and social contributions. As in many other countries, there is no single cause of the arrears problem. Rather, the persistence of arrears reflects a history of failed tax amnesties, poor collection practices, weak enforcement, restrictions on writing off uncollectible debt, a lack of capacity to deal with debtors’ inability to pay, and inappropriate penalty and interest structures. This cycle of amnesties and weak collections is self-perpetuating. To break it, a comprehensive solution that addresses the different specific aspects of the arrears problem is needed.

Below is a description of the recent tax amnesty programs in Turkey.

The “Tax Peace Plan” (Early 2003)

In early 2003, the Tax Peace Plan amnesty was introduced. The participation incentives were moderate: (1) arrears were recalculated by applying the change in the wholesale price index (typically below market interest rates) to overdue principal, (2) repayment was extended over 18 months, and (3) accrued interest was waived during the repayment period. In order to remain in the plan, participants were required to keep current with their tax liabilities.

The August 2003 Amnesty (Social contributions)

In August 2003, the government introduced a scheme to restructure the stock of social contribution arrears. The scheme’s main benefit was to introduce a 60-month repayment period. A 10 percent penalty was introduced (but not graduated over time), and compound interest rates were used instead of simple rates to encourage compliance.

The 2005 Amnesty (Social contributions)

In April 2005, a public receivables law was introduced in parliament, which provided for a number of blanket tax exemptions and amnesties. One of these related to social security arrears, the stock of which was estimated to have reached 6 percent of GDP. Although passage of this law was put on hold, it created an expectation of an amnesty, eroding compliance with social contribution collections.

To stem the decline in revenues and encourage tax compliance from taxpayers who wanted, but were unable, to comply with their tax liabilities, the authorities replaced the blanket amnesty program with a new program. The new framework provides for different payment schemes for different debtors, based on their ability to pay.44 This payment installment program applies only to arrears accumulated before March 31, 2005.

Effects and Assessment

  • Except for the 2005 amnesty for social contributions, these tax amnesties have raised little revenue.45 By substituting for proper collection methods, they may also have helped cultivate a culture of noncompliance.
  • Although some consider the Tax Peace Plan a success, the evidence is not clear. Of the 380,000 taxpayers who were accepted into the program, roughly one-third defaulted. Although TL 4.8 billion was collected out of an expected TL 6.6 billion, there was no monitoring of whether participants complied with their current tax or social security liabilities.46 These conditions could be waived, given the difficulty of calculating and enforcing penalties. Therefore, payments to the Tax Peace Plan may have come at the expense of creating new tax or social security arrears or increasing evasion.
  • The August 2003 social contribution restructuring recovered little and did not keep pace with the increase in arrears. For one of the social contribution agencies (Social Fund for Employees, or SSK), out of the TL 4.2 billion eligible for restructuring (almost all the SSK total), TL 2.8 billion was entered for the scheme, of which less than a quarter—TL 0.5 billion—was collected. For another social contribution agency, Bağ-Kur, applications to restructure TL 1.7 billion were received (compared with reported arrears of 10 times this amount), of which TL 0.85 billion was collected.47
  • Some have claimed that the August 2003 social contribution restructuring failed because its terms were less generous than the Tax Peace Plan, and that the participants expected that a more generous amnesty would be offered. However, it is not clear that the results of the 2003 amnesty were any worse than those of 2001 and 2002 (low amounts of arrears were applied for restructuring then). If the Tax Peace Plan’s terms were more generous, it is not surprising that it collected more revenue. But this does not mean it was more successful, because collections through this plan could have come at the expense of accumulating new arrears elsewhere.
  • From 2001 to 2003, penalties and interest were much less than the prevailing market interest rates (approximately 35 percent versus 80–100 percent). The incentive to borrow through the government was likely an important reason for the buildup of tax arrears during this period and suggests that an appropriate penalty and interest structure is crucial to avoid a situation where taxpayers accumulate high tax arrears.
  • The lack of attention to the importance of cut-off dates for taxpayers to sign up for amnesties undoubtedly contributed to the expectation of future amnesties and reduced the likelihood that they will help increase taxpayer compliance. The fact that rumors of tax amnesties began long before cut-off dates were announced likely discouraged taxpayers from remaining compliant with their liabilities. The timing of the cut-off date is crucial.
  • The experience of the 2005 social contributions amnesty highlights the importance of revenue administrations having the capacity to write off tax debts that are deemed to be unrecoverable, so officials can focus on those arrears that can be recovered. In Turkey, the social security agencies’ legal inability to do this hampered their ability to clean up their books and focus on the most recoverable arrears.
  • Although it is difficult to measure the effect of these tax amnesties on taxpayer compliance, it is notable that (except for the improved tax revenue performance that was experienced in 2005) tax revenue performance as measured by the tax-to-GNP ratio has been undynamic from 2003 to 2007 (projected). It has hovered around the same level since 2003, and is projected to be nearly the same in 2008 as it was in 2003 (Table 4). Social contribution collections increased significantly with respect to GNP between 2005 and 2006 as a result of the 2005 social contributions amnesty. However, whether the improved revenue performance will be sustainable in the medium to long term remains to be seen.
Table 4.Turkey: Tax Revenue Collection, 2003–08(In percent of GNP)
20032004200520062007 (Proj.)2008 (Proj.)
Tax revenue23.823.424.623.923.423.7
Social security contributions5.
Source: IMF staff.
Source: IMF staff.


Turkey’s repeated tax amnesty programs do not appear to have been successful either in raising net long-term revenue or in increasing compliance. These programs have been introduced independently of major tax policy changes or of fundamental improvements in the capacity of the tax administration to enforce collection of tax liabilities and contributions.

The Philippines


About 18 tax amnesty programs were introduced between 1972 and 1987; these amnesties concerned most taxes, including taxes on income, motor vehicles, goods, and property. In some amnesties, taxpayers with disputed assessments were also eligible. In addition to these tax amnesties, which were mostly announced through presidential decrees, the Bureau of Internal Revenue Administration (BIR), starting in the 1990s and continuing up until today, “administratively” mitigated or waived penalties and surcharges. This took place on a large scale, in principle as an incentive for taxpayers to notify the BIR voluntarily of previously undeclared income. The following are some examples of the amnesty programs referred to above:

  • October 1972. This was the first of many tax amnesty programs in the Philippines; the Marcos regime offered the amnesty to taxpayers a few weeks after the declaration of martial law. The purpose was to give tax evaders the chance to reform and be a part of the “new society.” The amnesty covered untaxed income for 1971 and earlier—earned from sources inside and outside the Philippines. Any taxpayer who took the amnesty was granted a waiver of unpaid income taxes and penalties related to the nonpayment. Civil and criminal liabilities imposed under relevant laws, including the Anti-Graft and Corrupt Practices Act, were also waived. Freedom from investigation and examination of the amnesty return was also offered. The amnesty arrangement required payment of 10 percent of the previously untaxed income in lieu of the income taxes, interest, surcharges, and penalties.
  • November 1972. An amnesty was offered to owners of untaxed or improperly taxed motor vehicles in exchange for payment of 30 to 50 percent of the taxes due. Similarly, an amnesty was offered to importers and other owners of untaxed goods. The persons in question were in turn required to pay a 25 percent tax on the value of the goods; the taxpayers were relieved of relevant penalties, and no questions were asked as to the origin of the goods. Another arrangement gave an opportunity to taxpayers with delinquent accounts and pending contested tax cases to settle their tax obligation without civil, criminal, or administrative liability. Under this program, taxpayers faced a 40 percent discount on their debts if payment was made on or before January 31, 1973, or a 20 percent discount if payment was made on or before February 28, 1973.
  • March 1973. An amnesty was offered to those who failed to declare their 1972 income accurately. The program targeted an increase in tax revenue compared with 1971. As such, the amnesty in general required the taxpayer to pay an amount at least equal to the income tax for 1971 plus 40 percent. The immunities and privileges offered under this amnesty were the same as under the October 1972 program.
  • January 1974. This program had a broader coverage than previous income tax amnesties because it also covered inheritances and gift tax liabilities. It extended the amnesty both to taxpayers who had already availed themselves of previous amnesties but had declared their income only partially, and to taxpayers who had not availed themselves of the previous amnesties. The amnesty required payment of 15 percent of untaxed income or wealth.
  • January 1975. This program was also related to income and wealth and was called the ultimate (i.e., last) tax amnesty. However, many more tax amnesties were implemented in the following years—so in reality an amnesty was offered every tax year until 1987. Some of these programs differed slightly from the previous programs because they had a somewhat narrower coverage—with regard to both tax types and beneficiaries. In addition, a common requirement of the amnesty programs was that taxpayers had to file a statement of net worth (wealth). Further, the authorities repeatedly attempted to impress upon taxpayers that after this “one-time tax amnesty,” the authorities would vigorously enforce the existing tax laws.
  • August 1986. This program targeted undeclared income and wealth and was declared a one-time tax amnesty covering the years 1981 to 1985. The Aquino government implemented this program shortly after it had taken over from the Marcos government during the so-called People Power Revolution. Despite the fact that in 1986 many considered that the numerous tax amnesties offered under the Marcos regime had seriously eroded taxpayers’ confidence in government, the Aquino government still decided to carry out its own version of a tax amnesty. The rationale for this amnesty was that taxpayers had been discouraged from paying their taxes, as a form of civil disobedience during the Marcos regime—particularly for 1981–85. The amnesty required that taxpayers (1) file a statement of net worth as of December 31, 1980, and as of December 31, 1985, and (2) pay taxes equivalent to 10 percent of the increase in their net worth from January 1, 1981, to December 31, 1985.
  • September 1986. Taxpayers with disputed assessments were allowed to settle their tax obligations by paying 30 percent of the assessed amount. Penalties and other liabilities related to these cases were also waived.
  • February 1999. The BIR implemented the Economic Recovery Assistance Payment program, which granted immunity from audit and investigation to taxpayers who had paid at least 20 percent more than the tax paid in 1997 for income tax, value-added tax, and other taxes.
  • October 2002. The BIR implemented the Voluntary Abatement and Assessment Program (VAAP), which was offered to those taxpayers found (through the BIR’s analysis of third-party information) to have understated their sales. The BIR informed the taxpayers of its findings and gave them a chance to correct previous returns voluntarily without risking severe penalties. This program is still in place today.

During the entire period covered above (the 1970s to the 2000s), tax amnesties were implemented in an environment characterized by a weak revenue administration and a culture of low tax compliance. Extensive efforts over the past 10–15 years to improve revenue administration, supported by substantial external assistance, have unfortunately not brought about significant or sustained improvements in enforcement of the tax laws. The revenue administration suffers from severe weaknesses, some of which would make the launching of tax amnesties a risky undertaking.

Effects and Assessment

Recent studies and a public debate in the Philippines on the suitability of tax amnesties to generate revenue offer interesting insights on this issue. For example, in the context of analyzing the appropriateness of introducing yet another tax amnesty, a 2003 study by the Philippine Congressional Planning and Budget Office (CPBO) on tax amnesty programs implemented between 1972 and 1987 concluded that the track record of past amnesty programs was poor. The study showed that tax amnesty programs generated only P 2.9 billion in a span of 15 years, which translates into an average collection rate of 1.6 percent of annual tax revenue (Table 5). The CPBO concluded that the BIR’s VAAP program of using third-party information to identify potential fraudulent taxpayers was likely to generate much more revenue than a new amnesty; this program had yielded approximately P 6.25 billion from October 2002 to January 2003. The CPBO proposed that the BIR be given enough time to test the efficiency of the VAAP and that consideration of the tax amnesty proposal be deferred. Some industry and legal groups opposed the implementation of the proposed tax amnesty of 2003, claiming that such a measure would result in moral hazard, that is, it would discourage the public from paying taxes, in expectation of the next amnesty.

Table 5.The Philippines: Tax Revenue Collection from Selected Tax Amnesties, 1972–87
YearTax CoveragePercent of Total Collections
1972Undeclared income/wealth, motor vehicles, nonfilers of income tax returns0.02
1974Undeclared income/wealth2.47
1975Undeclared income/wealth2.81
1981Undeclared income/wealth2.11
1986Undeclared income/wealth2.15
1987Undeclared income/wealth0.61
Source: Philippine Congressional Planning and Budget Office.
Source: Philippine Congressional Planning and Budget Office.

Although the department of finance generally supported the 2003 draft tax amnesty bill, it expressed reservations regarding the frequency of tax amnesties in the Philippines. It strongly recommended the inclusion of a provision in the amnesty legislation that would put in place a moratorium on the granting of future tax amnesties. It argued that tax amnesties, when offered too frequently, could be counterproductive. Tax amnesties could be interpreted as a sign of weakness on the part of the government and its enforcement mechanism, and might encourage taxpayers not to fulfill their tax obligations based on the expectation of yet another future amnesty. In a presentation that the Department of Finance made to the House Ways and Means Committee in relation to the 2003 draft bill, it pointed out that because tax amnesties had practically been a yearly event in the past, tax collections resulting from amnesties were relatively minimal except during dramatic turning points in the Philippines’ political history—for example, the declaration of martial law in 1972 and the 1986 People Power Revolution, when many had high expectations of positive change. As mentioned above, the 2003 draft amnesty proposal did not get sufficient political support and was thus never implemented.

Some amnesty programs implemented in the 1980s and 1990s required tax evaders simply to pay a certain percentage of their taxes due; other programs required that taxpayers, in addition to paying, also file a statement of assets, liabilities, and net worth as a part of the amnesty arrangement. This latter requirement was set in an attempt to create a long-term compliance effect. The statement was intended to provide the opportunity for the BIR to update its taxpayer database, broaden the existing tax base, and provide a foundation for establishing benchmarks for future tax assessment purposes. However, the statements were never used to facilitate tax administration, and the data more or less went to waste because the BIR had no capability to manage this information effectively.

As Table 6 shows, the ratio of tax revenues to GDP declined steadily between 1999 and 2004.48 It is likely that the combination of repeated tax amnesties and weak tax administration efforts have been important factors fueling Philippine taxpayers’ somewhat relaxed attitude toward tax compliance, resulting in a level of compliance that is now alarmingly low. Tax evasion is widespread; many businesses operate outside the tax net, and hundreds of registered taxpayers simply fail to file returns.

Table 6.The Philippines: Tax Revenue Collection, 1999–2005(In percent of GDP)
Tax revenue14.513.713.512.512.412.413.0
Source: IMF staff.
Source: IMF staff.


The Philippine tax amnesty programs have had several goals, including (1) to generate revenue; (2) to give noncomplying taxpayers the opportunity to comply with the tax law and to start with a clean slate without fear of penalties; (3) to facilitate tax administration, as tax officials would no longer have to go to the trouble of identifying, verifying, and prosecuting tax evaders; and (4) to trigger a spirit of reconciliation between government and taxpayers that could result in a stronger culture of tax compliance. However, the declining trend in tax revenue collections (in the 1999–2005 period) suggests that noncompliance rates have not fallen but have increased during this period. This suggests that the tax amnesty programs have not achieved their intended results, but instead are likely to have encouraged taxpayers to forgo or delay tax payments while speculating that another tax amnesty may be forthcoming. The result has been a continuous deterioration in taxpayer compliance that is likely to have been an important contributing factor to a decline in tax revenue over the years.

17It is likely that the U.S. states’ experience represents upper bounds in collecting revenue from tax amnesties because the states have a relatively low-cost and credible way to improve their auditing technology: they can cross-check data with the federal government (Stella, 1991).
18See, for example, Torgler and Schaltegger (2005) for an overview of the various tax amnesty programs across the world. Alm (1998) analyzes the Russian experience with tax amnesties.
19For more details, see “Commonwealth of Kentucky: 2002 Tax Amnesty Final Report,” Kentucky Revenue Cabinet, November 2003.
20Other features were (1) a short participation window (the program ran for 61 days starting from August 1, 2002), (2) widespread taxpayer eligibility (only taxpayers under criminal investigation or those subject to criminal litigation were not eligible), (3) inclusion of all state taxes, and (4) an extensive period of time covered (all tax evaded from the end of the previous tax amnesty coverage (December 1, 1987) up to December 1, 2001).
21Eighty-seven percent (20,510) of applications for amnesty were eventually accepted. Those that were not ultimately approved were rejected because the taxpayer failed to pay the liability in full.
22The new filers category includes both first-time registrants (87 percent of new filers) (i.e., people who have never filed any taxes with the state), and taxpayers who are already registered for some taxes but requested new accounts for other taxes (13 percent of new filers). The first-time registrants contributed $1.2 million (or less than 1 percent of total amnesty revenue).
23It is also interesting to note that taxpayers reported less tax evasion for 2001 (the year immediately prior to the 2002 amnesty). This could reflect several types of strategic behavior on the part of taxpayers, including the expected and pre-announced toughening of penalties for tax evasion in the post-amnesty era, which would be expected to increase compliance.
24For more details, see “State of Michigan Tax Amnesty Program 2002,” Office of Revenue and Tax Analysis, Michigan Department of Treasury, February 2003.
25Other features were (1) a short participation window (the program ran from May 15 to July 1, 2002); (2) widespread taxpayer eligibility (both individuals and businesses)—only taxpayers subject to a current tax-related court case or criminal investigation were not eligible; taxpayers with existing tax assessments or accounts receivable were eligible to participate; (3) inclusion of all state taxes; and (4) an extensive period of time covered (all tax evaded up to June 1, 2001).
26The 22,220 amnesty taxpayers filed an average of 3.3 returns each. Amnesty and interest payments averaged $1,243 per return.
27Excluding interest payments, 60 percent of the amnesty tax payments were for tax years after 1999, and another 25 percent for the tax years 1996 to 1998, and just 0.6 percent were for tax years before 1988.
28The 1993 amnesty also brought in some IR£75 million in respect of “straightforward” arrears of tax: tax that had been declared but remained unpaid. Effectively this was akin to the 1988 interest amnesty.
29Many taxpayers made multiple payments (for different types of taxes); it is not clear how many individual taxpayers were involved.
30Irish Revenue does not systematically publish estimates of noncompliance for the major taxes. However, data on total tax revenue collections and the ratio of tax arrears to total tax collection suggest that noncompliance has been falling steadily. Total tax receipts collected by Irish Revenue increased from approximately IR£12 billion in 1988 to approximately IR£55 billion in 2005, an increase of 358 percent over the 17-year period. Total tax debt as a proportion of total tax collection fell from about 38 percent in 1988 to about 3 percent in 2005. These trends are reflected in the figure in Appendix I.
31Based on the findings of a technical assistance mission from the IMF’s Fiscal Affairs Department, August 4, 2005.
32The amnesty initially lasted from November 2001 to May 2002 and covered capital held abroad before August 1, 2001 (Government of Italy, 2001). For discussion of the content of the amnesty, see Manzitti (2002).
33This penalty was waived if the funds were invested in Italian government securities.
34The projected structural revenue increase is based on the assumption that all repatriated capital remains in the form of financial investment and is subject to a final withholding tax of 12.5 percent.
35This impact is difficult to ascertain because the Italian economy was subjected to significant external shocks at the time.
36For example, assuming that repatriated assets were held abroad for an average of 10 years and that about 50 percent of these assets resulted from previous tax evasion, the revenue forgiven by the Scudo Fiscale amounted to 1.1 percent of GDP and the tax discount reached 65 percent of taxes owed.
37These arguments are supported by the findings of a 2005 task force that analyzed the major weaknesses of Italy’s public accounts. The task force concluded that recent amnesty programs have undermined tax compliance in Italy. The findings are discussed in Faini and others (2005).
38A similar tax amnesty was offered in 2003 yielding €0.6 billion in new tax revenue—which corresponds to about €24 billion of newly regularized offshore investment.
39Official statistics are not available showing noncompliance levels by tax in Argentina prior to 2000. However, statistics published by the National Agency of Public Revenue of Argentina in 2005 show that value-added tax (VAT) noncompliance increased from 27.9 percent in 2000 to 34.8 percent in 2002 (during the economic crisis), and then fell to 24.8 percent in 2004. Although the overall trend for these four years shows improved compliance with the VAT, these figures nevertheless suggest that noncompliance levels for this tax in terms of total tax collection are still relatively high by international standards (by way of contrast, the VAT noncompliance rate in Chile in 2004 was about 19 percent).
40For more details, see Salim and Vassallo (2005).
41For example, a 2004 law offered a tax amnesty to taxpayers in the simplified tax regime who had ceased their activity without formally communicating this to the tax authority. The tax arrears that appeared as though they had accumulated since these taxpayers ceased their activity (although according to the tax department they were still economically active) were waived. Thus, part of the problem with the accumulation of tax arrears could be attributed to an inaccurate taxpayer register for taxpayers in the simplified regime. Instead of solving this problem through administrative procedures aimed at updating the register and ensuring it contained accurate information, the government issued an amnesty.
42An amnesty introduced at end-2003 offered taxpayers in arrears a 50 percent reduction in overdue interest payments and a waiver of all financial penalties. Many taxpayers had accumulated arrears because of the high interest rates on overdue tax and the ensuing rapid accumulation of overdue interest payments. However, rather than more reasonable penalty interest rates being included in the tax legislation, an amnesty was offered.
43Based on technical advice provided by the IMF’s Fiscal Affairs Department, August 2004.
44The installment payment facility that was approved was tailored to ability to pay for only a reduced number of large debtors who kept good records. Other debtors with poor records and for whom it was difficult to ascertain capacity to pay were offered the amnesty.
45The 2005 amnesty raised sizable revenues, around 0.6 percent of GNP, from “collection of arrears from social security institutions.” A further 0.2 percent of GNP in collections of these arrears was projected for 2007.
46Prior to the Tax Peace Plan (2002), the authorities were collecting approximately TL 1.2–1.5 billion from tax arrears and accrued penalties and interest, clearly less than that collected through the Tax Peace Plan. Thus, viewed purely from the perspective of tax arrears collection, the Tax Peace Plan was successful.
47Though Bağ-Kur’s arrears stock appears to have been quite massive, a high proportion of this stock was not meaningful, because it related to contributors (especially farmers) who did not realize that they were members of Bağ-Kur.
48Factors such as trade liberalization and the nonindexation of excises have been important contributors to the declining revenue trend.

    Other Resources Citing This Publication