Tax Evasion in Developing Countries Term Paper (Development Economics) DSE According to Swiss Coalation of Development Organisations, harmful tax practices is leading to a loss of at least USD 50 billion a year for the Developing Countries. This amount is equivalent to the amount required by the World Bank and UNDP to achieve its Millenium Development Goals. It is also equivalent to the annual official aid given by the OECD countries to Developing Countries. And the tax burden is ultimately shifting from the rich to the poor. Introduction
After the opening up of the financial markets tax evasion and money laundering have come up as serious problems for macroeconomists. Even though tax evasion was always a concern for the economists it has now increased to a much larger extent,due to the fact that many of the new financial markets assit the tax evaders by providing them with an easier way of hiding their black money. This amount which would have been used for economic and social development purposes is used for either criminal activities or for riskier low-quality investments like gambling.
This money on entering the stock market also brings about a lot of instability. In a study done by Alex Cobham,for low-income countries ,shadow economy forms 32. 7% of the GDP and also has a potential tax revenue of 13. 7%. And therefore, by bringing down tax evasion a lot of funds can generated and used for more welfare oriented investments. Earlier Studies 1. Theoretical studies: The first classical model on tax evasion was made by Allingham and Sandmo in 1972 where an individual tries to maximise his expected utility from evasion of taxes.
The model considers a monetary fine along with a loss of reputation as a penalty and had static as well as dynamic versions of the same. It also included comparative static results with respect to actual income, tax rate , penalty and probability of detection concluded that probability of detection and penalty rate both or a combination of the two can be used for reducing tax evasion. But they assumed that fines are assessed on income evaded by the tax payer.
So, the affect of increase in taxes was unclear as the substitution effect (higher marginal tax rate makes evasion more attractive) and income effect (high tax rate lowers income) offset each other incase of risk aversion( as low income makes taxpayer less willing to take risk). However, in 1974 Yitzhaki showed that substitution disappears when fines are assessed on taxes evaded and not income evaded and solved the ambiguity related to effect of tax rate in Allingham-Sandmo model. After this the Allingham-Sandmo-Yitzhaki model was used as the benchmark for all other later models.
All these models assumed the tax payer to be corrupt but not the tax collector which may be true for the developed nations but not for the developing countries. However, the models by Becker and Stigler (1974), Rose-Ackerman (1975), Klitgaard (1989 &1995), Basu Bhattacharya and Mishra (1992) Mookherjee and Png (1995) and Nadeem Ul Haque and Ratna Sahay (1996) assume tax collector be corrupt as well and used the principle-agent problem to find the solution to tax evasion and corruption.
Among these the model by Nadeem Ul Haque and Ratna Sahay shows that by increasing the incomes of government employees more skilled individuals can be attracted to the job and also higher government wages also weakens the incentive compatibility condition of corruption, for the tax collector thereby reducing corruption and tax evasion. Generally all these models assume the tax payer to be dishonest but the model by Brian Erard and Jonathan S. Feinstein assumed that not all tax payers are dishonest and used an asymmetric information static game to build up their model and had results very dissimilar to the others.
The theoretical models made till now still have a lot of scope to come closer to the reality of corruption present in the developing countries. As most of the results depend on the kind of assumptions regarding who all are considered to be honest and dishonest it is difficult to choose and say which one is more appropriate. 2. Empirical Studies and general views on tax structure : An Analytical Framework 1. Related Work : This is an extension of the Nadeem Ul Haque and Ratna Sahay model. The original model consists of four sets of players :the government, the tax collectors, the auditors and the private sector firms.
Each private sector firm produces yi, on which the government imposes a uniform tax rate of t, such that the total tax accruing to the government is T= ? tyi = t? yi . So the tax paid by each firm is Ti = tyi . But the actual tax reported by the private firms is ? i = ? Ti , 0<? <1. There is perfect information between government and tax auditor as well as between the tax collector and the private firm. But there is imperfect information between the tax auditor/government and tax collector/private firm. And so , the private firm and the tax auditor engage in mutually beneficial agreement in which the firm underreports its tax as ? and gives the firm a bribe of bi for cooperating with it. Thus the firm gains Ti – ? i from tax evasion and because the firm and tax collector share in gains from tax evasion bi ? Ti – ? i has to hold. But the government is aware of this kind of cooperation between the firms and the tax collectors so it sets up institutional arrangements to monitor this kind of behaviour. It hires tax auditors who are capable of detecting tax evasion with a probability p, 0 ? p ? 1. This p is directly proportional to the skill of the auditor which in turn is dependent on his /her relative wage ? (wg/wp), where wg is the wage of the government employee and wp is the wage of the private sector employee. The assumption behind this argument is that , a higher wage will be able to attract more skilled people into the job of an auditor and so a more skilled auditor will be able to detect tax evasion and corruption more efficiently thus increasing p. And so, p is an increasing function of ? . p is also directly proportional and increasing function of the amount of tax evaded, as the chances of detection increases. And so, p = p(? , T- ? ) . The function p is assumed to be differentiable, increasing and convex in both arguments, i. . ?p/?? > 0 and ? p/? (T-? ) > 0 , where T = ? tyi and ? = ? ?Ti , which is observable to the government and so in a way p is a choice variable for the firms. After detection both firms and tax collector are fined and the penalty is larger than the amount cheated. The amount of penalty or fine charged is proportional to the amount of tax evaded . It is ? p ( Ti – ? i ) for the private firms , where ? p > 1 and ? g ( Ti – ? i ) for the tax collector , where ? g > 1. Based on the above mentioned details the firm maximized its gains from evasion by choosing ?.
And the gains from tax evasion were divided between the firm and the tax collector depending on their respective bargaining powers , which determined the bribe bi given to the tax collector. After maximization the amount of tax evasion ? i was found to be a function of the policy variables of the government. Therefore ? i = ? i( ? ,? p ,? g ,Ti ) and total derivatives with respect to ? , ? g , ? p ,were found to be positive. 3 This means that as the relative wages of the auditors is increased (probability of detection raised) or the penalties are increased tax collection improves.
But the effect of an increase tax rate and thus Ti is ambiguous as an increase in tax can have two effects one direct effect which increases tax collection and another indirect effect which increases tax evasion and so the overall effect need not be necessarily positive or negative. The most important results of the Nadeem Ul Haque and Ratna Sahay model are that by increasing the relative wages of government employees : tax collector and the auditor the government budget deficit 4 can actually be reduced as increase in tax collection might be higher than the increase in wage expenditure. . Criticisms : a) The implicit assumption of this model (also the other principle-agent models) is that the tax auditor is honest and the probability of detection only depends on his skills , is not very convincing. It is similar to the assumptions made in the initial models where the tax collector was assumed to be honest and only the tax payer was dishonest. These models also incorporate the same flaw of arbitrarily assuming some people to be honest and others to be dishonest .
It is very well possible that if the gains from tax invasion are too high the evader might actually make the auditor also collude with him/her as the amount at stake from higher gains is also higher . It is very unlikely that the evader would passively let the detection depend upon something exogenous but would rather endogenise the probability of detection which is possible by bribing the the auditor as well . And it is reasonable to assume this as corruption in low-income countries is not restricted to one level or a group of people but is present at multiple levels and overtime becomes a part of the system. ) The other assumption of probability of detection or audit being the same as punishment is also not that appropriate. In real life it is not always that if an individual is detected he / she will be automatically punished as unless it is proved in the court that the individual is guilty of tax evasion there is no effect of detection. As it is the case with most of the developing countries the judicial system is not very efficient .
Firstly, it takes quite a lot of time in giving the judgement and so a delay in judgement and always increases the benefit from evasion as the punishments are discounted more with delays and so loose some of their threatening power. Secondly, even after a lot of delays (and many a times due to delays) the evader may not be proved guilty due to lack of evidence(there is a possibility of corruption there as well). As a result of both these factors the effective probability of punishment, which is the actual probability that would be considered by the tax payer while making the decision of evading taxes goes down drastically.
And generally this is case in most developing countries where a large proportion of raids (most used method of detection ) does not result in punishment. Thus, the probability of punishment is the appropriate probability to be used in this model. And a simple application of the model without taking into account the effective probability may not give the desired results. c) The relative wage ? of the tax auditor has no relation with the number or magnitude of detection. It is assumed that an increase in wage will bring in more skilled workers and so automatically will increase detection.
But having a more skilled worker does ensure an incentive to detect also. Higher skill may imply higher ability to detect but not necessarily more will or incentive to detect. And so, it is equally important for there to be incentive along with skill to have better detection mechanism. Otherwise, plainly increasing the wages of the tax auditor might be a futile exercise which is one results of the Nadeem Ul Haque and Ratna Sahay model. So, there should be rewards for detection for the auditor to have an incentive to work more efficiently.
These rewards can be explained as the promotions which government officers get in real life. Therefore, successful detections (the ones for which punishment is given) should determine wages and not the other round. 3. The Model : In this framework, there are three players : the government, the tax payer or the private firm and the auditor. Here both the tax payer as well as the auditor are considered to be rational i. e. the auditor will be ready to take a bribe if maximizes his expected gains by doing so.
The tax collector has not been included as it firstly, does not make any difference to the working or the results of the model and secondly because if the the firm can bribe the auditor himself bribing the tax collector does not make any difference. The tax rate , each firms produce , and tax collected or reported are the same as in the original model t, yi, ? i respectively , which implies Ti and ? are also the same as before. One major departure from the original model is that now instead of one probability of detection there will be a combination of the probability of audit and the probability of punishment incase of an audit.
Thus, now the required probability p will be a product of the probability of audit p1 and the probability of punishment after the audit ( successful proving of tax evasion in the courts) p2 . So, p = p1 * p2 . The penalty imposed on the firm incase of punishment is the same ? p [ Ti – ? i ] , where ? p > 1. The firm is also assumed to be risk neutral. Another change is that, the auditor will now get a reward for successful detection of tax evasion. The reward shall be a proportion r of the amount evaded and so that the higher the amount of tax evaded the higher the incentive to detect.
If the reward will be amount irrespective of the amount of tax evaded there can be a tendency on the part of the auditor to make maximize the number of audits rather than maximizing the amount of tax collection. The reward should also be only for those audits in which there has been successful punishment, so that the auditor cannot claim reward on those audits where there was no tax evasion. In that case the auditor will audit the maximum number of people without getting any evidence for any of those audits. Therefore, there should be two parts of the remuneration of the auditor one the fixed , regular wage wg and so the relative wage ? nd the second , the variable component of the wage r ? i( Ti – ? i ) , i here implies number for all the firms which have been punished for tax evasion. This mechanism would ensure the maximum effort from the auditor without giving him/her any undue advantage. Working : A) Conditions for minimizing corruption – In this case any collusion between the firm and the auditor would take place only if the following two Incentive- Compatibility-Conditions are satisfied :- a) The gain from corruption should be higher than expected gain from no corruption for the firm. i. e. giving bribe to the tax auditor should be more beneficial than not giving. Ti – ? i ) – bi ? ( Ti – ? i ) – p1 * p2 { ? p [ Ti – ? i ] } ? p1 * p2 { ? p [ Ti – ? i ] } ? bi -(1) Equation (1) implies that the expected penalty from punishment which takes place only if evasion is first audited and second verified in the court should be greater than the bribe offered to the auditor. b) The gain from taking a bribe should be higher than the gain from the expected reward. This implies if the auditor is offered a bribe he would assess the chances of him getting the reward after auditing in comparison to the bribe he would get for not auditing: i ? p2 { r [ Ti – ? i ] } -(2) From equations (1) & (2) for both the Incentive-Compatibility-Conditions to meet simultaneously : ? p1 * p2 { ? p [ Ti – ? i ] } ? p2 { r [ Ti – ? i ] } -(3) (3) should hold . ? ? p * p1 ? r Therefore, if r > ? p * p1 then no corruption on the part of the firm or the auditor would take place. This means if that if an amount slightly greater than the entire penalty amount is given to the auditor there will be no incentive for corruption from both the parties.
B) Conditions for maximizing tax collection without corruption : The expected gains for the firm in case of no corruption are – Max (Ti – ? i)[( 1- p1 ) + p1(1 – p2)] – ? p(Ti – ? i)[ p1 * p2 ] ? i ?Max (Ti – ? i)[1- p2] – ? p [(Ti – ? i)( p1 * p2 )] ?i The first order conditions are : -(1 – p2) + ? p(p1*p2) + ? p * ? i * p2 * (? p1/?? i) = 0 [ as p1 is an increasing function of (Ti – ? i) or a decreasing function of ? i] ? ? i = (1 – p2) – ? p(p1 * p2) [? p * p2 (? p1/?? i)]
