I. Introduction
There is a common understanding among tax theorists and practitioners that more effective tax administration performance1 is associated with higher tax compliance, but what is the evidence of this relationship? This paper addresses the question using an empirical analysis based on a theoretical framework. With many tax administrations globally having invested in strengthening their institutional performance, particularly regarding the Value Added Tax (VAT), it is reasonable for them to expect the compliance gap would shrink. But quantifying the relationship between these two variables is not so straightforward, not least because neither tax administration performance nor tax compliance are easy to measure.
The role of the tax administration in tax compliance has been studied in the literature, but without addressing how the strength of the tax administration system as a whole translates into compliance outcomes (see Appendix I). Silvani and Brondolo (1993) and Agha and Haughton (1996) were the first to carry out a cross-country empirical analysis of the determinants of VAT compliance. Silvani and Brondolo (1993) correlate the characteristics of VAT design with the compliance gap in 20 countries, but they recognized that tax administration effectiveness was an important omitted variable in their model given the lack of data. Agha and Haughton (1996), in a similar analysis for 17 OECD countries, set out to represent the effect of tax administration through the size of its budget resources. Following this line of analysis a series of other studies have emerged which we use as the basis to develop our analysis, including Christie and Holzner (2006), Reckon (2009), Keen (2015), Das-Gupta et al (2016), Crivelli (2018), CASE (2020), and Butu (2021) (See Appendix I).
We use the standard Allingham and Sandmo (A-S) theoretical framework to analyze the correlation between tax administration performance and VAT compliance gaps (see Appendix II). We apply the theoretical model developed by Allingham and Sandmo (1972) and Sandmo (2005) to the VAT. The model’s main assumption is that a representative VAT payer decides to evade a given portion of his/her liability based on a compliance risk portfolio decision. We complement the model by adding a variable relating to tax morale: the degree of the taxpayer’s social acceptability of the tax system. 2 The advantage of using this well-known (if simple) compliance model is that it identifies the factors representing the tax administration's actions and performance, and it considers taxpayer behavior. The model mirrors closely the modern Compliance Risk Management (CRM) approach that tax administrations adopt when designing compliance strategies.
We measure tax administration performance using well-established assessment tools and estimate its impact on compliance gaps (the difference between what should have been collected based on the tax laws and what is actually collected). Our empirical model is based on a panel dataset consisting of data generated via the following tools: Tax Administration Diagnostic Assessment Tool (TADAT), Revenue Administration Gap Analysis Program (RA-GAP), and International Survey on Revenue Administration (ISORA). We use the TADAT dataset to calculate an average performance score ranging from ‘A’ (best) to ‘D’ (worst) based on an array of different tax administration characteristics and practices. For the VAT compliance gap, we use estimates by country and year obtained through a method we call the ‘reverse RA-GAP'. We use the ISORA dataset to obtain information on the tax administration’s budget and human resources. We also use datasets of macroeconomic variables, VAT rates, and international indexes from IMF and World Bank datasets. In some cases, we supplement this information with data from each country’s official MoF website. The resulting panel includes 111 countries—mainly EMEs and LIDCs—during the period 2010–2023.
Our main finding is that better tax administration performance is associated with a lower VAT compliance gap, showing that it does matter—and quite substantially. We estimated an elasticity of -0.7, so if a tax administration were to raise its TADAT score from 1.85 (close to a ‘D+’) to 2.32 (close to a ‘C+’), one could expect an increase of 0.6 percent to GDP in additional VAT revenue. If we also incorporate the fall in the CIT compliance gap induced by the decrease in the VAT compliance gap, one could expect an additional 0.7 percent of GDP in CIT revenue. Thus, the total effect would be approximately 1.3 percent of GDP in additional revenue from both taxes.
Based on our results we also establish that: Investing efforts in improving tax administration performance can lead to reduced noncompliance as a result mobilizing additional domestic revenue, although it may take time to generate such gains. Based on the data, we observe that achieving an improvement from 1.85 (close to a D+) to 2.32 (close to a C+) is not a short-term endeavor; based on the evidence from the sample of countries that have undertaken a repeat TADAT (about 30 countries) this could take an average of 5.8 years.
If the social acceptability of the tax system could be improved, with the tax administration (and other public agencies) playing an active role in fostering a national environment that promotes this objective, it will open another channel for reducing the compliance gap and increasing associated revenue. We estimate a semi elasticity of -0.13, which suggests that if the actions of the various government institutions can lead to an increase in this indicator from -0.62 to -0.57, this could raise revenues on the order of 0.04 percent of GDP, on average, by reducing both VAT and CIT compliance gaps.
The tax administration’s compliance strategy should take into account the structure of the economy given the latter’s role in determining compliance. In countries where the VAT is primarily collected at the border, customs administrations can help improve compliance by strengthening import controls and systematically providing revenue-related information to the tax administration; the latter can use this as a source of ‘third party’ information, for example, to check taxpayers’ compliance with their domestic VAT obligations. Tax administrations face greater challenges in economies dominated by the agricultural sector (given the prevalence of many small enterprises that are often informal and operate on a cash basis) and must apply tailored and robust compliance risk strategies to detect and address noncompliance effectively.
The rest of the paper is organized as follows. Section II describes the theoretical framework used to estimate the impact of tax administration performance on the VAT compliance gap. Section III presents the empirical model based on country panel data. Section IV concludes.
II. Framework
A. Theoretical Motivation Our approach is based on the standard tax compliance risk theory3 (Allingham and Sandmo (1973) and Sandmo (2005)). The theory considers a representative taxpayer who rationally decides to evade a portion G of his/her liability based on the financial benefits derived from such action (See Appendix II). With a certain probability, p, his/her noncompliance behavior can be detected and penalized by the tax administration, and with another probability, (1-p), it can go undetected, resulting in a net gain derived from the noncompliance action. Based on individual’s income level, I; the tax rate, t; the perceived probability of detection, p; and the fine rate, X, applied in case of being detected, the optimal level of evasion, 𝛤∗ , comes from maximizing the expected return of the risk portfolio over these two probabilistic scenarios.
𝛤 ∗ = 𝛤∗(𝐼,𝜏,𝑝,𝑋). (1)
The first-order condition predicts that both the fine rate and the probability of detection have a negative effect on the compliance gap, while the effects of income and the tax rate are indeterminate.4
We characterize the income level and the tax rate based on a representative business subject to VAT. In the model, the representative business produces a single product that is taxable for VAT purposes (see Appendix II). The business’s income corresponds to the true value added it generates, which also defines the base subject to VAT. We refer to this variable as the VAT-able base. For the tax rate, we consider the standard VAT rate applicable to the business’s single product.
The probability of detection is often represented in the literature by the tax administration’s expenditure (See Agha and Haughton (1993) and Feltenstein et al. (2022)). The probability of detection is associated with the volume of resources that the administration has available to enforce tax compliance with respect to the universe of potentially enforceable transactions or taxpayers. One could express this in several ways, e.g., the tax administration’s expenditure compared to GDP, the number of audits executed compared to the universe of potential audits, or the number of staff compared to the total population.
The fine rate is typically established in the tax law, but the ability to effectively apply it can also reflect the tax administration’s expenditure. In the tax law, fines usually depend on the severity and intent of the evasion. In monetary terms, statutory fines can be two, three or more times the amount evaded.5 But in practice, the tax administration's ability to apply fines effectively may be a more important determinant than the amount of the fine as per the tax law. Often, the effective application of the penalty regime implies a years-long effort by the administration to pursue the imposition of such fines through the courts. Thus, achieving effective penalty enforcement will depend more on the tax administration’s efforts than on the level of the penalties themselves. For this reason, we also use the tax administration’s expenditure to represent this determinant.
In our analysis it is not only the magnitude of the tax administration’s expenditure, 𝒆𝑻𝑨, that affects the probability of detection, p, and the actual fine rate, X, but also the effectiveness with which it spends its resources, 𝜼 (See Christie et al. (2006) and Reckon (2009)). This implies that two tax administrations with the same level of resources can generate different levels of probability of detecting noncompliance depending on their performance. For example, adopting a Compliance Risk Management (CRM) approach in many tax administrations has made it possible to increase the probability of detection without necessarily adding new resources or staff.6 Something similar can be observed in two tax administrations that operate with the same level of fine rates for VAT noncompliance but one is more effective in applying such fines.7 The following equations reflect the effect of the tax administration’s expenditure to GDP, 𝑒𝑇𝐴, and the administration’s effectiveness in applying its resources, 𝜂𝑇𝐴𝐷𝐴𝑇,8 to the probability of detection, 𝑝, and the fine rate 𝑋.
𝑝 = 𝑝(𝑒𝑇𝐴,𝜂𝑇𝐴𝐷𝐴𝑇) (2)
𝑋 = 𝑋(𝑒𝑇𝐴,𝜂𝑇𝐴𝐷𝐴𝑇) (3)
Other variables can facilitate or hinder the probability of detection and the imposition of an actual fine. ‘Facilitating’ variables include the application of VAT on imports, the existence of legal powers to control and sanction noncompliance, and the degree of simplicity of the tax legislation, the existence of third-party data (customs, financial, etc.) among others.9 The most recognized ‘hinderers’ are a large agriculture sector, special country conditions (Fragile and Conflicted-Affected States (FCS), Small Island) often associated with weak institutional capacity, the prevalence of small enterprises transacting in cash, etc. We test the effect of some of these in the next section.
Beyond tax enforcement variables, non-financial determinants of tax compliance are also found important under the broad umbrella of tax morale (See Alm (2019)). These may include the perceived value of government services that are financed via taxation, trust that the tax administration acts fairly, the social value of complying with norms, the cost of compliance10, and the government’s political legitimacy. We will use the concept of ‘social acceptability of the tax system’ (Diamond and Saez (2011)) to capture most of the non-financial factors affecting VAT compliance. 11 The variable 𝑎 will represent these factors as follows:
𝑎 = 𝑎 (𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑠𝑒𝑟𝑣𝑖𝑐𝑒𝑠,𝑡𝑟𝑢𝑠𝑡 𝑖𝑛 𝑡𝑎𝑥 𝑎𝑑𝑚𝑖𝑛𝑖𝑠𝑡𝑟𝑎𝑡𝑖𝑜𝑛,𝑠𝑜𝑐𝑖𝑎𝑙 𝑛𝑜𝑟𝑚𝑠,..) (4)