Effects of tax reform on Argentina's revenues

P.ONY RsseweI / 5 WORKING PAPFRS Countr Opera Latin America and the Caribbean Country Department IV The World Bank Septmber 1993 WPS 1192 Effects of Tax Reform on Argentina's Revenues Jacques Morisset and Alejandro Izquierdo In Argentina, changes in tax legislation, tax administration, and individual taxpayers' attitudes toward tax evasion improved tax revenues. Here is a method to measure how much. PaWyR RWaogPq tdiu=rnAfindiw.pin pui an d i .dachngf idme Ban amaffad &D1y *Airvew-ba.ldb.weudntuecbmy1dmbfWbnt8R.chinUAdeeu3iyOWmmLthiVdaCim so be anumW to t W=Id BDha ke Bmd ofDitZ. i mmgmor myof i4 _mbw aau,_  Country Opwldons WPS 1192 This paper - a product of the Country Operations Division, Latin America and the Caribbean, Country Department IV - is part of a larger effort in the department to understand fiscal reforms in Argentina. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Gerald Carter, room 16-020, extension 30603 (September 1993, 23 pages). Too often, a good tax policy proposal is consid- * The admin;strative dimension of tax reform ered sufficient to improve the tax system - too is at the heart of Argentina's recent fiscal little consideraticii is given to weaknesses in tax adjustment. Since 1991, tax effort is an average administration, perhaps because of measurement 80 percent higher than during the preceding problems. Analyzing legal and administrative (temporary) successful adjustment period (under measures and quantitatively evaluating their the Austral Plan). impact on tax revenues is generally arduous. *An efficient tax ad:;.nistr~tion and an Morisset and Izquierdo develop a simple improvement in taxpayer compliance levels approach to assessing how tax effort affects tax appear to precede rather than follow increases in revenues (performance). By "tax effort" they tax revenues. mean changes in tax legislation (except changes in nominal taxes), tax administration, and * Tax effort is influenced significantly by individual taxpayers' attitudes toward tax such macrovariables as GDP growth and infla- evasion. Changes in tax administration include tion, as well as by political (in)stability. It is increasing tax penalties, new technologies, and influenced less by such fiscal variables as administrative reform. altemative sources of financing. They measure tax effort as a residual: the * In Argentina, the sequence of the tax effort variations in tax revenues that cannot be ex- was, first, to broaden the potential value-added plained by changes in economic variables and tax base, and then to reduce tax evasion through tax structures. Using this approach, one can higher tax penalties and improvements in the easily identify factors that influence tax revenues basic functions of tax administration (inspection, over timne, and understand the behavior of tax audits, tax management, and personnel policy). revenues in developing countries, particularly where macroeconomic conditions are volatile. The authors apply this approach to Argen- tina; it can as easily be applied to other countries. Their main conclusions in this application: The Policy Research Working Paper Series disseminates the findings of work under way in theBank. Anobjectiveof the series is to get these findings out quickly, even if presentations are less than fully polished. The fndings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the Policy Research Dissemination Center  Effects of Tax Reform on Argentina's Revenues by Jacques Morisset and Alejandro Izquierdo 1 We would like to thank R. Bird, E. Gxerken, J. Hanson, and C. Kane for helpful comments on a previous version of this paper. We are very grateful to J. Vazquez-Caro and R. Newfarmer for providing us information on tax administration in Argentina and for valuable comments. Section 2.4 of this paper draws extensively on their work on Argentina.  Table of Contents Summary Introduction 1. A simplified Approach to Measuring Thx Effort , 2. An Empirical Analysis of Tax Effort in Argentina 2.1 Estimation of Tax Effort (1983-92) 2.2 The Causal Relationship between Tax Effort and Tax Revenuc. 2.3 Determinants of Tax Effort 2.4 The Administrative Aspect 8 3. Conclusion Bibliography Tables and Figures Appendix  Sumniar 1. The objective of this paper is to assess the contribution of tax effort to tax revenue performance. Tax effort, used in a somewhat unconventional way, includes changes in tax legislation (except the changes in nominal taxes), in tax administration and in individuals' attitudes toawrd tax evasion. Although the imnportaiace of administrative reform is widely acknowledged by those concemed with tax reform in developing countries, this issue has been relatively little explored in the economic literature, perhaps reflecting measurement problems. 2. Changes in tax administration, ranging from incrasing tax penalties to technological progress and administrative reforms, would require careful analysis of a myriad of legal and administrative relationships. It would be even more arduous to evaluate quantitatively the impact of these actions on tax revenues. An alternative approach is developed in this paper; tax effort is viewed as a residual; i.e. the variations in tax revenues that cannot be explained by the changes in economic variab'-s and in tax structure. The major benefit from this approach is that the factors that influen_e tax revenues can be easily identified over time, and provides a good understanding of tax revenues behavior in developing countries, particularly with volatile macroeconomic conditions. This approach has been applied to Argentina, but can be readily applied to other countries as well. 3. The major conclusions can be summarized as follows. First, the administrative dimension of tax reform is at the center of the recent fiscal adjustment in Argentina. Since 1991, tax effort is on average 80 percent higher than during the precedent (temporary) successful adjustment episode (the Austral Plan). Second, an efficient tax administration and an improvement in taxpayers compliance levels appear to precede rather than to follow increases in tax revenues. Third, tax effort is influenced significantly by macrovariables such as GDP growth and inflation as well as by the stability of the political environment, but relatively less by other fiscal variables such as alternative sources of financing. Finally, the sequencing of tax effort in Argentina was, first, to broaden the potential VAT base, and, then, to reduce tax evasion through higher tax panalties and improvements of the basic functions of tax administration: inspection and audits, tax management, and personnel policy.  INT(?UCTION In the economic literature on taxation, policy reveals little concern for weakness in tax administration. A good tax policy proposal and technological progress is thought to unambiguously improve the tax system. However, administrative constraints may prevent the establishment of an optimal tax system, particularly in countries suffering from a scarcity of trained administrators. "In shon, there may well be too much preoccupation with what to do and too little attention to /how to do it" (Bird, 1992, p.189). The objective of this paper is to assess the contribution of changes in tax administration to the evolution of tax revenues. The virtual abs._nce of studies on this issue may reflect both conceptual and measurement problems. Changes in tax administration, ranging from iicreasing tax penalties tc technological progress and administrative reforms, would require careful analysis of a myriad of legal and administrative relationships. It would be even more arduous to evaluate quantitatively the impact of these actions on tax revenues. For these reasons, an alternative approach is developed in this paper. The effect of economic variables such as inflation and GDP growtl:, and changes in tax structure on tax revenues is distinguished from the impact of changes in tax administration and in the taxpayers behavior. These two last factors, defined as tax effort, are viewed as a residual; i.e. the variations in tax revenues that cannot be explained by the changes in economic variables and in tax structure. The major benefit from this approach is that the factors that influence tax revenues can be easily identified over time. The approach is applied to the case of Argentina. One of the prominent features of the Argentine tax system is the high volatility of total tax collection. Abrupt changes from year to year reflected increases or decreases in the level of economic activity, in the rate of inflation, tax rates, in taxpayer compliance levels, and tax administration efficiency. The approach developed in this paper will place the administrative dimension of tax reforn at the center of the success of the current adjustment program. During the 1989-92 period, tax revenues incrmased by more than 200 , rcent in Argentina. Using simple procedures, we will show that changes in tax effort are likely to have preceded rather than followed tax revenues variations. We will also attempt to identify the economic and political variables that influenced tax effort as welJ as the channels used by the Argentine Government to increase tax effort. The paper proceeds as follows. In Section 1, we identify the main factors that inAuence tax revenues fluctuations by developing a simple approach to measuring tax effort. In Section  2 2, this approach is applied to the case of Argentina. In Section 2. 1, the contribution of different factors, including tax effort, to tax revenues fluctuations are measured over the 1983-92 period. Section 2.2 di--usses the issue of the causal relationship between tax effort and tax revenues. In Section 2.3, we attempt to identify empirically the variables that have influenced tax effort in Argentina over the last decade. In Section 2.4, we focus on the microeconomic aspect of the tax reforms implemented during the 1989-92 years in assessing to which extent the improvement in tax effort was due to a reduction in tax evasion or in tax exemptions. We also review changes in the enforcement regime and loopholes, and then tax administration itself. Finally, Section 3 contains our conclusions. 1. A SIMPLFIED APPROACH TI MEASURING TAX EFFORT The purpose of the approach is to identify the main factors that influence tax revenue fluctuations. In particular, we need to elaborate on the meaning of tax effort in the context of the paper, as this concept is used in a somewhat unconventional way. As a starting point, tax receipts collected by the tax administration can be defined as: (1) Tt = tB, where T, is tax revenue collected at time t by the administration, t, the average nominal tax rate, and B, the tax base. Since the inflation rate may be high and very volatile, the Olivera-Tanzi effect should ba included in equation (1). As suggested by Olivera [1967] and Tanzi [1978], tax revenues are influenced negatively by the rate of inflation given that taxes are collected with a certain lag. Following these authors, tax revenues can be expressed as follows (2) T, = t, [ B*,/[(l + Og)(l + 07r,)]J where B*t is the potential tax base, e the estimated delay in collection (days/month ratio), & the growth rate of the tax base and x, the inflatior. rate. Equation (2) states th, t tax r'venues depend on the average tax rate and the potential tax base adjusted for the Olivera-Tanzi effert Using this equation, we can determine the factors  (3) OlnT, = CMnt, + n1nB*, - Cln(I + eg) - nln(l + 07r) The percentage change in tax reverut; depends on the percentage changes in the theoretical tax base, on the average nominal tax rate, and on the Olivera-Tanzi effect. Equation (3) can be used to derive changes in the theoretical tax base, which is the only variable not directly observable. Moreover, assuming that the variations in the tax base are attributable to changes in economic activity or in tax administration, we can write the following equation: (4) nInB*, = ndnY, + nlnE, where CMnY, is the percentage change in the tax base owing to var.ations in economic activity. Substituting equation (4) into equation (3), the percentage change in tax cffort equals (nlnE): (5) CAnE, = n1nT, - nlnt, + C-ln(I + egj + cln(1 + er,j - n1nY, The percentage change in tax effort is therefore defined as the change in tax revenues that is not explained by va-.ations (i) in nominal tax rates; (ii) in the Olivera-Thnzi effect; and (iii) in the tax base due to changes in economic activity. The variable nLnE, is assumed to reflect important changes in tax legislation (except the changes in nominal taxes), in tax administration and in individuals' attitudes toward tr-- evasion. In Section 2.4, the variations in nTnE, due to changes in the tax base are distinguished from the other effects.2 While the results of such a simple accounting approach should be interpreted with care, they nevertheless offer a framework in which to appraise comparative tax effort over a period of time. 2 See footnote 10, for further details.  -4-: 2. AN EMIRICAL ANALYSIS OF TAX EFFORT E ARGENTINA 2.1 Estimation of 2x Effort (1983-92) rTe approach developed in the precedent section was applied to the case of Argentina over the 1983-92 period. After the poor performance of ihe Argentine tax system during the 1980s, tax revenues increased significantly from 12.7 percent of GDP in 1989 to 22.5 percent of GDP in 1992. One of the major achievements has been unamtiguously the surge in VAT revenues from 0.6 percent of GDP in the second quarter of 1989 tl 9.2 percent of GDP in the fourth quarter of 1992 (Figure 1). While part of this increase was due to more favorable macroeconomic conditions, improvements in tax administration and tax legislation also contributed to this positive evolution. Tl assess the contributions of these diff- rent factors on tax revenues, the percentage variations of th,e main components derived from equation (5) were calculated during the 1983-92 period on a monthly basis. Without loss of generality, we concentrated on the value-added tax (VAT), equivalent to about 60 percent of total tax revenues.3 While the complete results are presented in the Annex, the results obtained by using this methodology are illustrated in Table 1 for the stabilhzation periods June 1985-Septermber 1986 (Plan Aus.ral) and March 1991- December 1992 (Convertibility Plan). VAT receipts increased significantly during both episodes, but the monthly tax effirt was 80 perc,.nt higher during the Convertibility Plan than during the Austral Plan. In contr. st, if the decline in the inflation mte through the reversal of the Olivera-Tanzi effect accounted for an important part of the transitory success of the Austral Plan (explaining 30 percent of the total tax revenue increase), this effect was almost negligible during the Convertibility Plan. It has to be recognized, however, that about 12 percent of he increase in VAT receipts observed during the Convertibility program was due to the increase in the general VAT rate from 16 percent to 13 percent in March 1992. The evolution of the tax effort index (1983 = 100; see technical appcndix) over the 1983- 92 period is illustrated in Figure 2. During 1987-89, the deterioration in tax effort by about 75 percent was a major reason for lower tax revenues in Argentina. The capacity to administer 3 Note, however, that the irapact of the tax structure reform (e.g.elimination of distortionary taxes such as import taxes or stamp tax) on the efficiency of the tax system is excluded from this analysis.  - 5 - efficient taxes was eroded by inattention to management and systems development, frequent legislative changes and the imposition of new levies greatly complicated the work of the General 'Tax Board (DGD), resulting in the accumulation of inconsistent bureaucratic procedures. The change of Government in July 1989 produced an unprecedent improvement in tax effort --716 percent between the second quarter of 1989 and the tourth quarter of 1992. Tax effort increased particularly at the beginning of 1990 and during the first two quarters of 1992, in smuite of a temporary decline during the first quarter of 1991. As discussed in greater detail in Section 2.4, these improvements occurred in stages, beginning with the broadening of the tax base in Febnrary and November 1990. Losses in tax receipts owing to the Olivera-Tanzi effect also contributed to the fluctuations in tax revenues (Figure 3). As expected, the negative i.npact on tax revenues was significant during periods of high inflation, reaching its highest piunt during the two hyperinflations of 1989-90. The negative impact of inflation was also quite important during the months preceding the Austral Plan (June 1985). Not only was the domestic inflation rate extremely high, but the collection delay was estimated above 35 days (see Duran 1987). The fiscal reform package of mid-February 1990 required VAT and income-tax payment; to be made within 10 days, thus reducing the collection lag. 2.2 The Causal Relationship Between Tax Effort and Tax Revenues The administrative dimen'ion of the tax reform -xplains to a large extent revenue increase since March 1991. In absence of such effort, the increase in tax revenues observed during the Convertibility Plan would have been limited to 34 percent --much lower than the observed increase of 108 percent or even the increase in VAT collection registered during the Austral Plan. The purpose of this section was to determine whether tax administration reforms preceded or followed variationq in tax revenues. The geneally accepted opinion is that an increase in tax revenues generated by a good tax policy proposal improves the efficiency of tax administration. Tb determine the causal relationship between tax effort and tax revenues, we used the well-known Granger test with an optimal distributiun lag including 9 lagged variables.4 The results are summarized in Table 2. The hypothesis of Granger causality from tax effort toward 4 Corresponding to a -'hite-noise error distribution.  -6 - tax revenues cannot be rejected because the parameters of lagged tax effort appear jointly- significant at a 10 percent level (as shown by the F-stalistics in lhble 2). On the other hand, we did not find Granger causality from tax revenues to iax effort. Reg,1rdless if this test defines properly the concept of causality, it is important for at least two reasons. First, it indicates some prediction capacity of changes in tax effort in forecasting future variations in tax revenues. Second, the absence of Granger causality from tax revenues to tax effort allows us to reject the hypothesis that variations in tax revenues precede changts in tax effort. 2.3 rm inants of Tax Effort Since changes in tax effort are likely to precede rather than follow the variations in tax revenues, it is imporant to determine the economic and political variables which can aff6 changes in tax effort. Bird (1992) and Richuptan (1987) argue that macroeconomic variables are important in explaining taxpayers cormpliance and the efficiency of tax administration. However, the influence of macroeconomic variables on tax effort (or tax evasion) has been relatively little explored in the economic literature. The behavior of taxpayers can be viewed as ar. attempt to adjust their satisfaction or lack of satisfaction with government services and the macroeconomic environment, but the existing literature has only focused its attention on the explanation of the behavior of taxpayers in terms of micioeconomic factors (see Richuptan (1987) for a review of this literature). Assuming that risk-averse taxpayers want to maximize their expected utility (Allingham and Sandmo (1972)) or their expected income (Srinivasan (1973)), t6- effect of microeconomic factors --including tax rates, tax base, the probability of being detected and penalized, and the size of the penalty-- on tax evasion is discussed.5 However, this approach is difficult to test empirically because most of the explanatory variables are not directly observable. rax effort is also influenced by the fiscal authorities's decisions (as well as by the bargaining process between these authorities and taxpayers). One popular approach to understand the fiscal behavior of the public sector is t' ssume that it reflects the actions of a 5 Other factors such as the characteristic of the population --age, sex, educational background-- have also been explored.  set of public decision makers (e.g. Heller (1975)).6 Within this framework, tax effort would positively respond to an increase in public expenditures because the Govemrnments would have to adjust its buidget constraiut. Alternatively, additional sources of financing will reduce incentives to increase tax revenues. Finally, the efficiency of tax administration and the. taxpayers con.pliance levels ire closely related to the stability of the political system. Foll wing the above arguments, the empirical analysis of the responsiveness of tax effort has been limited to three classes of explanatory variables.' First, we tested fiscal variables: real public expenditures, the variation in mral net extemal public financing ((lnF), the variation in MI (nlnM) as a proxy of public monetary financing, an;l a dummy variable (DUMI) used to test the assumption that the prohibition of moneta%y financing during the Convertibility Plan exerted a positive impact on tax effort. Second, we introduced macrovariables: GDP growth (GROWTH) and inflation (INFL); and, finally, indicators of political stability: the number of months between each change in the office of the Minister (POL) and a dummy variJV.e (DTJM2) capturing the change of Administration in the third quarter of 1989. OLS regressions were carried out over the 1983:I-92:IV period, the most significant results are reported in Table 3. Additional comments can be found in th'- text.8 Vi; expressed the variables in percentage variation rat..r than in levels to eliminate the trend, and used lagged explanatory -ariables because these variables and tax eifort are likely to be jointl, determined. Although it is difficult to obtain stable relationships in a country like Argentina, the overall estimated results appear quite satisfactory as summarized by the explanatory power of the 6 Basically, public decision makers are assumed to maximize their utility taking into account uses of public resources and alternative sources of financing such as effort in taxation and borrowing. The maximization of this utility function with respect to current policy variables subject to the public sector budget constraint predicts that tax effort is related positively to public expenditures and negatively to other sources of funancing. 7 The analysis remains of course partial since changes in tax effort partially reflect the reality of political power and mnultiple aspects of the society such as the degree of corruption and the degree of civic conscience. 8 The procedure used in this paper differs from other studies in that the dependent variable is tax effort rather than total tax revenues. Using tax effort, the influence of macroeconomic variables through the Tanzi effect and changes in the potential VAT base is controlled for, implying more reliable tests of the impact of externaJ factors.  - 8 - regressions (adjR2) and the white noise behaviour of the residuals tested for autocorrelation (Lagrange multiplier test), heteroskedasticity (due to squares of the regressors) and normality (Jarque-Bera test).9 Tax effort appears significantly influenced by the political and economic environment, but little by other fiscal variables. The response of tax effort to past inflation appears to be positive and significant. This result is somewhat surprising because, from the taxpayers' point of view, an increase in the inflation rate is perceived as a policy failure that would encourage future tax evasion. However, it may indicate that the destabilizing impact of inflation on the public sector's budget during the 1980s'