Effects of electronic cash registers on reported revenue

Tax administrations around the world embrace digital transformation to enhance tax compliance. A notable example is the introduction of mandatory electronic cash registers (ECR)—certified electronic devices designed to register and calculate transactions at a point of sale. Motivated by its potential to process and monitor transactions in cash-intensive businesses, 16 OECD countries have already made certified ECRs mandatory by 2020.¹ Mandatory ECRs have also been raised by the European Commission as a way to curb under-reporting of value-added tax (VAT).² This has led to a recent initiative on “VAT in the digital age”, including recommendations on mandatory ECRs (European Commission, 2022b).

The introduction of ECRs reflects a broader shift towards digitized business transactions, with Electronic Invoicing (E-invoicing) being its closest counterpart.³ E-invoicing involves the electronic creation, transmission, reception, and management of invoices, typically using standardized digital formats. This method streamlines the exchange of invoice data between suppliers and buyers in electronic form, eliminating the need for traditional paper documentation. While E-invoicing primarily targets business-to-business transactions, ECRs are relevant for business-to-consumer transactions; however, both provide tax agencies with enhanced capabilities to detect and prevent VAT fraud.

Nevertheless, technological advancements in consumption taxation are a double-edged sword. On one side, these advancements can improve the monitoring and enforcement of tax compliance among firms, particularly in countries with limited state capacity (Fan et al., 2018; Bellon et al., 2022; Juniult, 2023). Conversely, such changes frequently provoke behavioral adaptations in the impacted businesses that may diminish the anticipated benefits (Agrawal & Fox, 2021). In addition, the efficacy of such measures in reducing VAT fraud largely hinges on their influence on audit effectiveness and the monitoring capacity of tax agencies. The overall impact of the implementation of mandatory ECRs is therefore an empirical question.

Despite the widespread adoption of mandatory ECRs, there is surprisingly limited availability of reliable firm-level data regarding their impact. Recent evidence from various developing countries suggests that the impact of ECRs can vary significantly depending on factors such as the implementation process and local context (Ali et al., 2015; Awasthi & Engelschalk, 2018; Fjeldstadt et al., 2020). An exception to the lack of evidence on the impact of ECRs in developed countries is a policy report by Lovics et al. (2019), which focuses on Hungary. The report shows that the introduction of ECRs led to a significant increase in reported revenue of 23–35% in the retail, accommodation, and food service sectors.

We contribute to this literature by studying the effects on tax filing behavior of a reform in Sweden that enforced businesses that handle cash payments to use a certified ECR. Under the legislation, all transactions made at the point of sale, whether by cash, debit, or credit card, were required to be recorded on a certified ECR. Before the reform, businesses were not required to use a specific type of ECR to record their revenue, nor were they obligated to issue a receipt unless specifically requested by the customer (Skatteverket, 2013). The Swedish reform not only facilitated the electronic storage of pertinent tax information, enhancing auditors’ ability to identify instances of tax evasion, but it was also accompanied by an increase in on-site tax inspections.

Our study benefits from access to unique administrative data containing information on high-frequency (monthly) output VAT for all Swedish firms that reported the use of ECRs to the Swedish Tax Agency. Our identification strategy exploits the fact that we can track firms both before and after acquiring ECRs at different points in time, providing robust insights into the causal effect of ECRs on reported revenue. Sweden is a country that shows high trust in the Tax Authority (Statskontoret, 2024) and a high level of tax compliance by small and medium-sized firms (DeBacker et al., 2015).⁴ On one hand, high tax morale and trust in the government imply that businesses in Sweden may be more willing to comply with tax regulations, including the use of certified ECRs. On the other hand, low initial tax evasion may also contribute to a modest effect on reported revenue.

We find that the implementation of an ECR by firms results in an immediate increase in reported revenue of 2.7–4.3%. Nevertheless, the impact appears to be temporary. Within a few months, reported revenue exhibited no substantial deviation from the levels witnessed before the implementation of the ECR. The results are robust to two distinct difference-in-differences (DiD) estimators that are typically applied in settings with staggered treatment timing: a standard two-way fixed effects (TWFE) estimator and a multi-period DiD estimator following Callaway and Sant’Anna (2021). The observed temporary increase in reported revenue is evident across virtually all sectors impacted by the legislation, with hairdressing, restaurants, and retail businesses being the most prevalent.

To establish a causal interpretation of our results, we demonstrate that the immediate increase in reported revenue after the acquisition of the ECR is preceded by a period of similar trends in reported revenue, covering a duration of up to one year. Moreover, we rule out the possibility that changes in the underlying economic conditions might have influenced the observed shift in reported revenue. To be specific, we demonstrate that there is no discernible alteration in the total wage cost of the firms, suggesting that they did not encounter significant changes in their economic environment around the time of the ECR implementation.

Our interpretation of the temporary nature of the effect is that firms have discovered novel methods of under-reporting their tax base. In a survey conducted by the Swedish Tax Agency, which targeted business owners affected by the ECR legislation, the majority of respondents viewed the reform as a favorable development that made it harder to evade taxes. However, some respondents pointed out that firms quickly adapted to the new regulations and found alternative ways to evade taxes (Skatteverket, 2013). This perspective is shared by employees at the tax authorities who work with auditing firms’ compliance with the ECR legislation. For instance, instead of canceling recorded revenues on their previous ECRs, firms now refrained completely from registering certain revenues on the new ECR.

Our main finding of a small and significant, but short-lived, impact on reported revenue underscore the mixed evidence presented in previous studies concerning the effectiveness of improved information exchange in mitigating tax evasion, which have mainly focused on third-party reporting.⁵ For instance, Slemrod et al. (2017), Adhikari et al. (2020, 2021) demonstrate that the effect of increased third-party reporting of credit card transactions on small business tax compliance is relatively modest on average, but more significant in business-to-consumer industries and among firms with initially weak tax compliance. Slemrod et al. (2017) recommend that information reporting be targeted specifically at groups suspected of significant tax evasion and those with large shares of income subject to such reporting. Adhikari et al. (2022) discovered that the increase in reported revenue following third-party reporting of credit card transactions in the United States was partially offset by increases in reported expenses. Carrillo et al. (2017) observe that most firms in Ecuador failed to respond to a threat of third-party reporting and conclude that governments face constraints in enforcement policies, even with access to information.

Our results also align with insights from an expanding body of literature that investigates the impact of audits on tax compliance. While audits exert an immediate mechanical effect by curbing tax evasion, their long-term impact is theoretically ambiguous. Audited firms may positively adjust their expectations of future audits (through Bayesian updating), fostering lasting compliance. Conversely, taxpayers might perceive audits as decreasing their likelihood of future scrutiny (referred to as the “Bomb crater effect” by Mittone (2006). Empirical evidence on the behavioral outcomes of audits varies. Some studies, such as those by DeBacker et al. (2018), Advani et al. (2023), suggest lasting positive behavioral effects (i.e., increased compliance), while Best et al. (2021) find no discernible impact in a large-scale randomized audit study in Pakistan. Furthermore, Gemmel and Ratto (2012) provide evidence indicating that compliant taxpayers subject to audits may subsequently decrease their compliance, in line with the bomb crater effect, while they find that non-compliant taxpayers increased their compliance.

The remainder of the paper is structured as follows: Section 2 presents the institutional background regarding the Swedish cash-register reform. Section 3 describes the data utilized in the analysis, while Sect. 4 outlines the empirical approach employed. The empirical results are presented in Sects. 5 and 6 concludes.

The cash register reform that we study in this paper was implemented in 2010. Prior to the reform, businesses could use any type of cash register to keep track of their revenues and were not obliged to provide a receipt unless requested to do so by the customer. This made it relatively easy for businesses to manipulate registered revenue or not report revenue at all (Prop. 2006/07:105). As in many countries, tax evasion among Swedish businesses in industries that handle large volumes of cash, such as retail, hospitality, and hair care, was widespread and considered a threat to the legitimacy of the tax system (SOU 2005:35).

The new legislation required businesses handling either cash or credit/debit card payments to use a certified ECR. A certified ECR should have a manufacturer declaration and a special control unit, a black box, accessible only by the staff at the Swedish Tax Agency. As part of the legislation, businesses were mandated to report their acquisition of a certified ECR to the tax authorities. However, small businesses that handle low volumes of cash were exempt from this requirement. At the time of the reform, the threshold for low cash balances was set at SEK 170,000 per year (1 USD \(\approx\) 8 SEK). Certain industries were also exempt, including the taxicab industry. The vast majority of firms purchased an ECR (as opposed to leasing one) with an average investment cost of approximately SEK 23,000 (Skatteverket, 2013).

Within two years of the reform, there were around 75,000 active firms with a certified ECR.⁶ Overall, the firms that acquired an ECR due to the reform correspond quite well to the specified target industries, i.e., small firms that rely heavily on cash transactions and whose customers are individual consumers with low interest in getting a receipt (Skatteverket, 2013). In the industries with the highest ECR density, between 60 and 70% of all firms have a certified ECR, including restaurants, hairdressers, and small-scale retail stores.

There were two additional elements of the cash register reform with potential implications for tax compliance. First, firms always have to print and offer the customer a receipt, independent of whether the customer asked for a receipt or not. Second, the reform gave the tax authorities enhanced audit rights, including unannounced inspections.

Within three years of the implementation of the new law, the tax authorities conducted more than 100,000 on-site inspections. The purpose of the majority of these inspections was to ensure that businesses complied with the new ECR requirements. Businesses not complying with the law were fined by SEK 10,000, and if the company once again failed to comply with the law within a year, a fine of SEK 20,000 was charged.⁷ These inspections were proportionally distributed across sectors depending on the sector’s share of the total number of ECRs. A recent study by Swedish Tax Agency does not show any signs that these inspections lead to improved compliance with regulations (Skatteverket, 2022). We therefore do not anticipate that the potential effects of ECRs will be confounded by the impact of these audits.

Due to the high demand for ECRs at the time of the reform, delivery times were often long and unpredictable for firms. For this reason, inspected firms without an ECR were not fined if they could verify that they had ordered an ECR. From July 2010, the authorities fined all non-complying firms. In 2010, 500 fines were implemented compared to 2,900 fines in 2012. Initially, the Tax Agency mainly fined firms for not having a certified ECR. Later on, fines tilted more towards reporting errors, such as not registering revenues or providing receipts (Skatteverket, 2013).

We use a staggered DiD approach to estimate the causal impact of ECR adoption on reported revenue among Swedish firms. The staggered treatment timing comes from the time variation in the adoption of ECRs across firms, following the new legislation in 2010 that enforced firms that handle cash payments to use a certified ECR. Our models are estimated using monthly VAT data at the firm level, with a symmetric window of 25 months around the treatment month.

Two-way Fixed Effects: We first evaluate the impact of ECR adoption using a standard TWFE model. We estimate both a static model and an event-study model that allows for dynamic treatment effects. In the static TWFE model, we estimate a single indicator measuring the average effect of implementing an ECR. The model specification takes the following form:The dependent variable, \(y_{i,t}\), is the log of reported revenue of firm i at month t. Firm-fixed effects are captured by \(\nu _i\) and month-fixed-effects by \(\delta _{t}\). The indicator \(D_{i,t,l}\in \{0,1\}\) captures the treatment status of firm i at month t being l month away from the first treatment. If a firm i has an ECR in month t it is treated and \(D_{i,t,l}=1\); otherwise it is \(D_{i,t,l}=0\). The first month of treatment is captured by \(l=0\). All firms are treated at some point between September 2009 and December 2010 and treatment is absorbing, i.e., \(D_{i,t,l}=1\) for all \(l\ge 0\). Our main interest lies in the parameter \(\lambda\) which captures the change in reported revenue after the introduction of the ECR. The parameter \(\lambda\) captures, in general, a weighted average of treatment effects.

The event-study version of the TWFE estimator allows for dynamic treatment effects. The purpose of this model is to examine whether the potential changes in reported revenue that follow from the adoption of an ECR are permanent or temporary. We estimate the following equation:Again, \(y_{i,t}\) denotes log reported revenue of firm i at month t, the indicator \(D_{i,t,l}\in \{0,1\}\) captures the treatment status of firm i at month t being l month away from the first treatment, and \(\nu _i\) and \(\delta _{t}\) denote firm- and month-fixed effects, respectively. The coefficients of interest are the \(\lambda _{l}\) that capture monthly reported revenue between 12 months before and 12 months after the adoption of an ECR. The month prior to the ECR, \(l=-1\), is excluded so that the estimate for each month is relative to the revenue level in that month.

Multi-period difference-in-differences: Recent econometric work raises concerns about the causal interpretation of two-way fixed effects parameters when there is treatment effect heterogeneity. Such concerns apply both to static TWFE estimators and dynamic event study models (Sun & Abraham, 2021). We therefore estimate the average treatment effect on the treated using the alternative estimator proposed by Callaway and Sant’Anna (2021). We measure group-time ATTs by comparing the expected change in the log of reported monthly revenue of firms that receive the ECR in a specific month with the respective change for firms that are not yet treated, i.e., have not acquired an ECR until then.

The group-time ATTs are then aggregated into a single parameter measuring the average treatment effect on the treated. More precisely, we implement the doubly robust DiD estimator based on inverse probability weighting and OLS (Callaway & Sant’Anna, 2021). This parameter is comparable with \(\lambda\) in the TWFE estimation in Eq. (1) but accounts for the heterogeneity of treatment effects across different groups (e.g., Goodman-Bacon, 2021; Baker et al., 2022). Similar to the TWFE estimation, we are interested in understanding the dynamic group-time ATT, which examines the heterogeneity of treatment effects in relation to the duration of exposure to treatment.

Given our sample restrictions, we cannot estimate treatment effects for the 11th and 12th months after the implementation of the ECR when we use the multi-period DiD model because there are too few observations in the control group of firms that have not implemented an ECR yet. Therefore, we will estimate the effects for the 10 months following the introduction of the ECR. In addition to estimating the effects for the 10 months following the introduction of the ECR, we also estimate the pre-treatment effects for the 10 months preceding the treatment.

We have shown that the introduction of mandatory ECRs in Sweden resulted in a short-term increase in reported revenue. We ascribe this response to companies anticipating greater costs associated with tax evasion. In our case, the expected costs can depend on multiple factors, such as the likelihood of being caught for tax evasion, the severity of the penalties imposed, or the efficacy of the monitoring mechanisms in place. According to survey data, most business owners believe that the tax authorities were able to monitor firms more efficiently due to the new legislation, which is a view that the tax authorities themselves endorse. Additional evidence supporting the claim that the reform enhanced tax compliance can be observed in an evaluation of the extent to which firms had to amend their revenues following a review by an auditor. The evaluation indicated that the audits conducted after the reform resulted in relatively minor increments in the reported amounts in comparison to those recorded before the reform (Skatteverket, 2013).

The introduction of ECRs made it risky to engage in fraudulent activities when customers paid with credit/debit cards, as all such transactions were now registered. However, it remained possible to circumvent the cash register entirely by paying with cash and not recording the purchase. Furthermore, in 2012, shortly after the reform, Sweden introduced Swish, an electronic payment system enabling instantaneous money transfers between subscribers. Consequently, customers could now conduct transactions outside the purview of cash registers, even without using paper money. In line with this, survey responses from business owners indicated that firms quickly adapted and identified alternative methods to evade taxes (Skatteverket, 2013). This illustrates the dual nature of technological advancements in combating tax evasion. Although facilitating easier information exchange and enhanced monitoring capabilities, such advancements also provide new avenues for swift and covert transactions.

While our findings may offer valuable insights for policymakers in other developed countries, it is crucial to contextualize the short-lived effects we observed. The reform under examination took place in 2010, a time when cash transactions still held considerable sway in Sweden, constituting 39% of all point-of-sale transactions (Riksbanken, 2022).¹⁴ This implies that our study captures a period where cash remained widely used, predating the significant decline observed in more recent years.¹⁵

In international comparison, Sweden belongs to a group of countries known for their higher tax morale, trust in the government, and relatively low incidents of tax evasion. Indeed, countries that mandate ECRs typically grapple with more substantial VAT tax discrepancies compared to Sweden (see Table 12). Among the countries that have implemented mandatory ECRs, Hungary presents a stark contrast to Sweden, having one of the largest tax gaps. Variations in tax gaps and underlying institutional features could account for the pronounced differences between our findings and those documented in the policy report by Lovics et al. (2019) on Hungary’s adoption of ECRs. It is, however, noteworthy that the reduction in the VAT tax gap over the last decade has not been markedly more pronounced in countries that have adopted mandatory ECRs compared to those that have not, as indicated in Table 12. To gain a more comprehensive insight into the influence of ECRs and other point-of-sale systems on tax compliance in developed nations, evidence from a broader range of countries is needed.

To ensure effective tax compliance, policymakers should adopt a comprehensive approach that goes beyond the introduction of ECRs. Many new and sometimes innovative approaches developed to support the formalization of point-of-sale transactions will have little impact on the shadow economy if applied without accompanied enforcement (Casey & Castro, 2015). Such complementary measures could include enduring tax inspections, third-party reporting, and improved information sharing between government agencies. In addition, by requiring businesses to use certified ECRs, tax authorities can better monitor all transactions, regardless of the payment method used, and detect instances of non-compliance.