How can informal interactions and bargaining affect tax compliance and enforcement?

In low-income countries (LICs), governments often struggle to effectively provide basic amenities such as education and healthcare. Consequently, many aspects of the economy are partly organised informally through interactions among citizens or between citizens and bureaucrats. For example, local public goods provision or credit markets. 

Tax collection is another important area where the state's capacity is typically low. While informal interactions between taxpayers and tax collectors are usually associated with detrimental corruption and thus lower tax collection, they can also make taxpaying smoother by discussing the tax process with taxpayers.

In my job market paper, with Eliya Lungu (from the Zambia Revenue Authority - ZRA), I empirically and theoretically establish bargaining over taxes between taxpayers and tax collectors as an important mode of tax compliance and enforcement. For the case of Zambian firms, we show that such bargaining (i) is a prevalent mode of tax compliance, (ii) can rationalise puzzling patterns in the administrative tax data, which are inconsistent with conventional models of tax compliance, and (iii) that bargaining over taxes is welfare improving in a low state capacity environment.

“Irrational” tax-filing?

In Zambia, as in most sub-Saharan African countries, small firms account for the vast majority of taxpayers. Yet, evidence of their tax compliance behaviour is limited. We analyse data on small firms from 2015-2021 in Zambia and find two examples of seemingly “irrational” behaviour. 

1. Firms bunch above tax schedule discontinuities

Zambia had a turnover tax for small businesses - a simplified tax system that is based on a firm’s revenue and not profits (which are often harder to calculate). During 2017-2018, the turnover tax schedule featured a linear tax rate and several tax brackets in which tax payments would drastically increase between the brackets. 

For example, if a firm generated revenue at the upper end of the first bracket, at ZMK 4200 in revenue, the tax liability is ZMK 36. If they earn one ZMK more, their tax liability jumps to ZMK 225. The better choice for firms would then be to pay less taxes, and so they would be better off disclosing revenue that is just below the tax bracket threshold. As illustrated in Figure 1, this leads to a big drop in net-of-tax turnover once the firm discloses one ZMK more and crosses the tax bracket threshold. The upper panel of Figure 1 shows that there is a dominated range of turnover in red above the threshold. This shows that all firms with turnover in the red field would be better off having a slightly lower turnover than ZMK 4200. Therefore, we would expect the distribution of firms to accumulate (“bunch”) below the threshold, but we would expect no firm to be just above the threshold. In contrast, the lower panel of Figure 1 reveals strong bunching above the threshold. The bunching shows that most firms exceed the thresholds by only tiny amounts.

Figure 1: Net-of-tax turnover around the threshold versus distribution of tax returns

2. Firms bunch at round payment amounts 

We also find strong and sharp bunching at turnover amounts, which imply rounding of tax liabilities, for example, tax liabilities, which are multiples of 10, 50, or 100. Before the tax brackets were introduced in 2017, the Zambian turnover tax schedule featured a flat rate of 3%. We can, therefore, clearly distinguish bunching at round number tax liabilities from the well-known phenomenon of bunching at round numbers of the taxable income itself. Round liabilities often imply particularly odd amounts of turnover. For example, we observe tax returns accumulating strongly at a turnover of exactly 3333.33 or 6666.66, which - at a tax rate of 3% - imply liabilities of 100 and 200, respectively. This suggests that the tax payments are not necessarily based on a firm's actual turnover.

Ultimately, we show that firms that are rounding payments more often are significantly more likely to bunch above thresholds. This suggests that both patterns might stem from the same underlying channel of tax collectors bargaining with taxpayers.

Firms discuss and agree on tax payments with tax collectors before filing

To learn more, I designed a survey to get a qualitative overview of how tax compliance works for small firms and, with the support of an independent survey firm, conducted 517 in-person interviews.

Most strikingly, the survey revealed that discussions with tax collectors over tax payments before filing tax returns are widespread. About half of all firms admitted that engaging in or considering such discussions was very common. In a follow-up question, 36% stated that such discussions serve the purpose of “finding an agreement” on what should be paid. Additionally, 57% stated that “clarifying their tax liability” was the purpose. These findings suggest that firms are bargaining over tax payments with tax collectors.

Bargaining over taxes can rationalise the empirical patterns

The survey indicated that what we consider to be turnover data may be more nuanced than it appears. Rather than representing true turnover, the data shows the tax amount decided upon by taxpayers and tax officials. This explains the seemingly ‘irrational’ behaviour of firms.

To see the argument, it is useful to consider bunching at round number tax liabilities first. This bunching pattern aligns well with the idea that tax returns represent the outcomes of bargaining. First of all, it explains the focus on the actual payment as the salient amount when filing taxes. Further, a large literature, starting with Nobel laureate Thomas Schelling, has highlighted that round numbers serve as focal points in bargaining situations. Thus, the strong accumulation of tax returns at turnover amounts which imply round payments may indicate that these are bargained and agreed upon payments. For example, when bargaining, it appears that two parties would likely settle on a payment of ZMK 100 instead of ZMK 99. 

Against this background, we explain how bargaining can rationalise the fact that firms bunch above notches – where they could increase net-of-tax turnover substantially by reducing declared turnover by a small amount. The key explanation is that notches restrict the set of payments over which can be bargained. When the tax schedule features jumps in tax liability at certain amounts, these jumps effectively introduce regions of payment that cannot be reached anymore. To see this, one can consider the example of the 4200 ZMK threshold in the Zambian case. When declaring 4200 ZMK as turnover, the payment will be ZMK 36. But when declaring a bit more, ZMK 4201, the payment will be ZMK 225. Thus, it is not possible to implement a payment between 36 and 225. When firms and tax officials are bargaining and settle on a payment within this interval (such as 50, 100, 150), the agreed-upon payment will end up either below or above that interval. In the tax data, this will create bunching on both sides of the threshold.

Implications for policy and welfare

One might ask the intuitive question of whether it would not be better to abandon any interactions between tax collectors and taxpayers to avoid corruption. However, our study makes the case that corruption is unlikely to be the main purpose of such interactions. Instead, we show theoretically that bargaining can lead to net tax revenue gains for tax authorities and reduce uncertainty in low-information environments. This welfare improvement occurs particularly when the costs of auditing or verifying tax liabilities are high, which depends on the level of state capacity. Our model thus explains why bargaining is more prevalent in low-income settings than in high-income economies, where widespread third-party information reduces the need for such negotiations.