ADMINISTRATION AGREEMENT BETWEEN THE EUROPEAN COMMISSION ON BEHALF OF THE EUROPEAN UNION AND THE INTERNATIONAL BANK FOR RECONSTRUCTION AND DEVELOPMENT TSI Project 20LT09 Micro Enterprises and Self-employed Tax Regulatory Assesment for Removing Hurdles to Growth Lithuania (EUROPE AND CENTRAL ASIA) Report Assessing the Impacts of Tax Optimization and Bunching in MEs and Self-employed and Legal Entities Responses to Size-based Tax Rates in Lithuania Output 1 & 2 December 2022 Project carried out with funding by the European Union in cooperation with the European Commission’s DG REFORM 1 DISCLAIMER This document was produced with the financial assistance of the European Union. The views expressed herein can in no way be taken to reflect the official opinion of the European Union. This report is a product of the International Bank for Reconstruction and Devel- opment/The World Bank. The findings, interpretation and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of the World Bank, the European Commission or the Government of Lithuania. The World Bank does not guarantee the accuracy of the data included in this work. Copyright Statement The material in this publication is copyrighted. Copying and/or transmitting portions of this work without permission may be a violation of applicable laws. 2 Contents 1 Preface 4 2 Introduction 7 3 Context and data 8 3.1 What can we learn from the experience of other countries? . . . . . 9 3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 4 Bunching responses to the change in average tax rate 13 4.1 Quantifying firms’ responses . . . . . . . . . . . . . . . . . . . . . . 14 5 Alternative response: changes in reported profit margins 19 6 Recovering elasticities using bunching behavior 21 7 Dynamic incentives: does the notch affect firm growth? 24 8 Conclusion 29 References 31 A Bunching Estimators 35 A.1 Revenue Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 A.2 Cost Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 B Robustness exercises 38 C Additional figures 43 3 1 Preface These reports correspond to the Outputs 1 and 2 of the TSI Project 20LT09: Micro Enterprises and Self-employed Tax Regulatory Assessment for Removing Hurdles to Growth in Lithuania. The two reports assess the impacts of tax optimization and bunching in micro enterprises and self-employed and legal entities and their responses to size-based tax rates in Lithuania, including the extent of the problem and drivers of tax optimization and bunching. To carry out the analysis the project team reviewed existing legislation and regulations to understand the overall context and the resulting incentive structure, it analyzed tax and other necessary data by the taxpayers to document existing trends and behaviors. As part of the project the team also analyzed relevant international experience and country cases to help benchmark the most relevant aspects of the system. Reports analysing the size and effects of tax optimisation, bunching and income shifting The reports identify the most important factors driving tax optimization and the provisions that are not aligned with international best practice that may lead to inefficiencies in the tax system, lack of neutrality and potential loopholes that impact on taxpayer behaviour, including on self-employed and MEs. The report on legal entities studies their response to the existence of differential tax rates based on firm size. While special treatments are intended to provide support for entrepreneurs and small and medium enterprises (SMEs) they may also provide the wrong incentives for some firms who may try to remain small in order to take advantage of the special treatment. The report documents the existence of a static response to differential tax rates in the short run and provides suggestive evidence that these effects extend to the medium- to long-term. The report on Personal Income Taxes and the incentives to shift income in Lithuania provides new empirical facts about personal income taxation (PIT) in Lithuania, with a focus on the implications of a schedular tax system for equity and efficiency. Different tax rates depending on the source of their income and differentiated rates encourage tax optimization and undermine the efficiency and equity of tax systems. The report documents the importance of self-employment income sources and how they vary according to income and focuses on owner- managers of small partnerships and they behaviour to minimize their tax burden. Separately, the project team has developed an analytical microsimulation model (part of Output 1) that can assist policy makers and stakeholders in simulating the taxpayer behaviour resulting in the existing optimization and bunching. The mi- crosimulation tool was used for comparisons with selected peer countries to simulate 4 possible effects of changes in the tax structure, tax parameters, or legal provisions related to tax optimisation and bunching. Data limitations, including added complications due to the COVID-19 pandemic that limited on-site missions did not allow the project team to collect the necessary data for estimating the impact of tax optimisation and bunching on overall produc- tivity and growth of MEs and self-employed using a macro-micro economic model (using the input-output matrix) differentiating by type and size of economic players in the most relevant sectors in the economy to estimate the economic productivity related to the size of economic units and their contribution to the economic out- put. The team documents however the existing practices of tax optimisation and bunching as well as other responses to regulation on growth, both for the ME and self-employed sectors, pointing to economic opportunity costs resulting from tax- payer practices attributed to the underlying incentives in the tax and regulatory systems. Micro level analysis on firm level productivity and identification of tax optimisation and bunching scenarios was impeded by lack of detailed data on the different cost variables at firm level, precluding the team from doing projections of productivity gains at the firm level. The reports have been produced by the World Bank team led by Mr. Alberto Leyton, Lead Public Sector Specialist (co-Task Team Leader) and Mrs. Cristina Savescu, Senior Economist (co-Task Team Leader). The authors of the reports are Thiago De Gouvea Scot de Arruda (Economist) and Pablo Garriga (Economist). The World Bank team would like to acknowledge and thank Mr. Vytas Adomaitis in the Innovation Agency of Lithuania and Mr. Vaidas Navickas, Adviser to the Prime Minister for their guidance, feedback, advice and hospitality, as well as coun- terparts in the Ministry of Finance, the Ministry of Economy and Innovation, the Ministry of Social Security and Labour, and the Central Bank of Lithuania for their feedback. The team would also like to express its gratitude to officials in the Depart- ment of Statistics - Edita Baltuše, Aleksandra Golubovič, Gita Literske, Vadimas Ivanovas, Darius Abazorius and Jonas Bačelis - for hosting the World Bank team in their secured office and for useful guidance in navigating through the data in the Palantir platform. Finally, the team would like to thank Kestutis Lisauskas, Si- mona Poceviciute, Akvile Laugalyte from Ernst Young for their support and advise for the interpretation and understanding of the specificities of the Lithuanian tax system. This report should be read together with the report on Outputs 4: Recommen- dation Report Analyzing the Size and Effects of Tax Optimization and Bunching with a Microsimulation Tool. This final version of the report has benefited from nu- 5 merous comments from Lithuanian authorities, as well as the European Commission and World Bank officials. 6 Corporate responses to size-based tax rates in Lithuania 2 Introduction This report studies how legal entities in Lithuania respond to the exis- tence of differential tax rates based on firm size. These types of policies are usually justified on the grounds of providing support for entrepreneurs and small and medium enterprises (SMEs). However, they may also provide the wrong in- centives for some firms who may try to remain small in order to take advantage of the special treatment. We document the existence of a static response (i.e., firms respond to to differential tax rates in the short run) and provide suggestive evidence that these effects extend to the medium- to long-term. Corporations in Lithuania face two possible tax rates on their income: a standard 15% rate and a reduced 5% rate for companies with less than ten employees and declaring turnover below €300,000.1 This sudden in- crease in average tax rates when firms declare slightly higher turnover—also known as notch —generates strong incentives to reduce declared turnover.2 To illustrate the incentives, consider a firm with turnover of €299,900 and profit (taxable income) of €15,000. That firm faces a 5% tax rate and will pay €750 in taxes. If the same firm increases its sales by €101 and declares a turnover slightly above €300,000, it now faces a 15% tax rate on all of its profit and, keeping profits constant, would face a tax liability of €2,250. In other words, a very small increase in turnover leads to a very large (€1,500) increase in tax liability. Firms respond to the incentives that are present in the tax schedule by bunching below the €300,000 threshold. This behavior is documented in Figure 2: it shows the distribution of firms according to their turnover with a distinctive accumulation (or bunch ) of firms reporting turnover exactly below the threshold. In the remainder of the document we quantify the bunching response and analyze its economic implications. 1 Our main analysis will focus on the period 2015-2020, after the adoption of the Euro. Similar incentives existed between 2011 and 2014; in the appendix we show that the same results hold for this period. In addition, firms in their first year of operation and below the same levels of employment and turnover are granted a zero tax rate on their first year of operation. 2 Kalanta (2021) provides an overview and previous explorations of the differentiated tax rates for small corporations. 7 Figure 2: Empirical density of turnover (2015 - 2020) Note: This figure presents the empirical density of turnover for firms in the period 2015 - 2020. Bins are €2,000 wide. 3 Context and data The design of a tax system with differential tax rates based on firm size usually is justified by the goal of providing support for entrepreneurs and SMEs (since these often respond for the majority of employment) and/or encouraging investment. For this purpose, governments usually define size-dependent thresholds meant to limit the beneficiaries of the policy (beyond the threshold the benefit disappears). In practice, it is commonly observed that firms will tend to optimize their behavior to locate at these policy thresholds. Bunching has been extensively documented the economics literature;3 here we provide a brief overview of studies that have analyzed similar situations in other countries. We then outline the data used for our own analysis. 3 Kleven (2016) provides an extensive discussion on the economic theory and applications of bunching responses to notches. 8 3.1 What can we learn from the experience of other coun- tries? Multiple studies document similar policy designs and convey the same message: taxpayers are keenly aware of large changes in incentives gen- erated by the tax system and will be strategic in their response, altering their behavior to avoid facing higher tax rates. Exploring the same kind of increase in average corporate tax rates as a function of turnover in Costa Rica, Bachas and Soto (2021) document many of the same behavioral responses seen in Lithuania: firms bunch below the threshold for preferential rates and report lower profits when facing higher rates. Kleven and Waseem (2013) also document similar behavior from personal income tax filers in Pakistan: when facing large increases in tax liability in arbitrary thresholds, agents bunch below them. Londoño-Vélez and Ávila-Mahecha (2019) show taxpayers bunching behavior in response to a notch on wealth taxes in Colombia. The extent to which these responses are real responses (i.e. changes in economic activity) or misreporting (changes in declared behavior, but not in real activity) is less clear, although all these studies suggest that at least in part misreporting plays an important role. Furthermore, it is worth noting that similar changes in incentives based on arbitrary thresholds (notches) exist not only on taxation, but in settings as diverse as car fuel regulation (Sallee and Slemrod 2012), electricity pricing (Ito and Sallee 2014) and retirement decisions (Manoli and Weber 2016). Similar to the behavioral response to taxation, agents also react to these notches in other settings and optimize their behavior in response to non-linear design of rules. The evidence on the effects of tax policy on entrepreneurship The role of corporate tax rates on the creation of business is a long- standing one and not entirely settled. Djankov et al. (2010) use cross-country data to document a negative correlation between effective tax rates and business density. Survey evidence from small and medium enterprises (SMEs), on the other hand, often point to other barriers (such as financial constraints and regulatory burden) as much more relevant to business decisions (Ravšelj, Kovač, and Aristovnik 2019; Wang 2016). This is a complex question to fully answer since taxation can affect entrepreneurship through several channels. Gordon and Li (2009) provide an extensive theoretical discussion of how the tax system can affect the decision of individuals to start businesses. Measuring the effect of taxes on entrepreneurship is 9 particularly hard because differential tax rates between corporate and non-corporate entities might drive choices of organizational form: a reduction in the corporate tax rate might drive entrepreneurs to incorporate, but that does not mean that new economic activity is being generated, just that income is being shifted from unincorporated activities to incorporated ones. Differential rates between wage- and self-employment income can also drive individuals’ decisions, either by making real economic decisions or by relabelling income (Saez, Schoefer, and Seim 2019). Overall, Gordon and Li (2009) conclude that the most important tax provisions to encourage risk-taking by corporations is beneficial treatment of losses (e.g. allowing it to be deducted from payroll taxes). More recent studies have focused on the distinction between low- and high-impact entrepreneurs. Definitions are not consistent across studies but high-impact entrepreneurs are typically defined as those more likely to file patents, have higher survival rates, higher number of employees, etc. Taking into account this distinction, the empirical evidence suggests that higher personal income tax and corporate income tax reduce the number of high-impact entrepreneurs. Chetty et al. (2019) and Akcigit et al. (2018) show that increases in the personal income tax have substantial effects on reducing the number of patents. Denes, Wang, and Xu (2019) studies the staggered implementation of angel investor tax cred- its in 31 U.S. states from 1988 to 2018. They find that tax credits increase the number of angel investments and average investment size. However, additional investments flow mostly to lower-quality startups that are launched by less expe- rienced entrepreneurs. Their findings suggest that state-level investor tax credits are ineffective in promoting high-impact entrepreneurship. More broadly, the con- sensus seems to be gravitating towards differentiating which type of entrepreneurs are the ones that actually contribute to economic growth or employment. General tax incentives may affect the selection of entrepreneurs taking advantage of such incentives, generating unintended consequences on the economy (Shane 2009). Supporting early-stage entrepreneurs might be more effective than supporting small ones. Using data for firms in the United States, Sterk, Sedláček, and Pugsley (2021) show that initial conditions matter for the future employment growth of firms. This is consistent with other evidence that not all entrepreneurs expect their firms to grow, bring new ideas to the market or create jobs (Hurst and Pugsley 2011; Abbring and Campbell 2005). If the policy goal is to encourage high- impact entrepreneurs who will generate growth, targeting those initial conditions may be a more effective way to encourage high-impact entrepreneurs than personal or corporate income tax rates. 10 Evidence on the effects of tax policy on investment and firm growth There exists ample evidence that firm investment responds to changes in the tax base (Ohrn 2019; Ohrn et al. 2022; Cummins, Hassett, and Hubbard 1994). Ohrn (2018) finds that a 1 p.p. decrease in effective tax rate in a program to encourage domestic manufacturing in the US increased investment by more than four percent. Zwick and Mahon (2017) show that accelerated depre- ciation provisions in the US were also very effective in increasing eligible capital, particularly for small firms. Bilicka (2020) shows that the investment of Canadian firms respond strongly to a tax reform that changed cash flow availability. On the other hand, Harju, Koivisto, and Matikka (2022) estimates null effects of a corporate tax cut in Finland on the average level of investment for affected firms. Evidence on the impact of broader tax cuts such as dividend tax, in contrary, often point to null results on firm investment. Yagan (2015) finds that the 2003 dividend tax cut in the United States did not increase the investment of affected firms, while payout to shareholders increased substantially (Chetty and Saez 2005). In France, Matray and Boissel (2020) estimate that a large increase in dividend tax increased investment, since firms started distributing less profits and investing the retained profits. In Sweden, Jacob (2021) finds weakly positive effects on investment when dividend tax was reduced, concentrated in firms that were liquidity constrained. These results are consistent with the so called "new view" on corporate investment, emphasizing that new investment is often financed out of retained earnings and therefore changes in the cost of external finance should not matter for investment decisions. More in line with the "traditional view" of corporate investment, Moon (2022) studies a cut in capital gain taxes in South Korea and finds strong investment responses, with particularly large effects on firms that were likely to be cash constrained. More broadly, governments often evaluate whether changes in the tax system will affect economic growth. In a recent review of 42 studies, Gechert and Heimberger (2022) find that corporate tax cuts have very limited, if any, effect on growth. Their key result is that, on average, studies find that a 10 percentage points (p.p.) reduction in corporate income taxes increase GDP growth by 0.2 p.p. However, there is evidence of publication bias: imprecise estimates showing null effects are less likely to be published. Correcting for that, the authors cannot reject the hypothesis that these tax cuts have zero effect on economic growth. 11 3.2 Data We use microdata on the universe of limited liability corporate taxpayers4 between 2015 and 2020 to characterize firms’ responses to the existence of different corporate income tax rates. In Table 1 we present key descriptive statistics for the data. The entire panel is composed of almost 600,000 firm-year observations, with between 90,000 and 105,000 filers per year in the more recent periods. We highlight a key change in the composition of limited liability firms in the period.5 In 2015 over 90% of them were joint stock companies (UAB) and 7% were small partnerships - the remaining entities are a mix of agricultural com- panies, cooperatives and a range of public institutions. Between 2015 and 2020, the share of small partnerships increased by 12 p.p., with an equivalent decrease in the share of private companies. Geographically, 80% of firms are from the three main regions of the country—Vilnius, Kaunas and Klaipeda. As in most countries, the distribution of firms is characterized by a long right-tail of large firms and a majority of small ones. While the average turnover (gross revenue) in 2020 was close to €1 million, the median was below €50,000 and over 15% of firms reported zero turnover. Approximately 20,000 firms (or 20% of those filing corporate income taxes) report turnover above the €300,000 threshold that defines larger firms (those not receiving the preferential 5% income tax rate). Across years, the average tax liability is quite stable at approximately €13,000 (median below €1,000). The same dispersion is observed for number of employees: the average number is 62 but the median is only 2 employees.6 4 These are entities filing the form PLN204. Sole proprietorships, general partnerships and limited partnerships that file form PLN204A are not included on this study. 5 That change in the nature of corporate entities is likely a result of incentives in the personal income tax system, which provides benefits for owners of small partnerships. This illustrates the important links between PIT and the choice of corporate form. 6 While the median number of employees is stable over time, the average increases steeply from about 11 in 2015 to over 60 in 2020. That is driven by an increase in the number of very large firms: the largest firm by number of employees in 2015 has approximately 45,000 employees and only 5 firms have more than 10,000 employees. In contracts, the largest firm in 2020 has almost 100,000 employees and almost 100 firms declare 10,000 employees or more. 12 Table 1: Descriptive statistics 2015 2018 2020 All years Firms’ characteristics Joint Stock Company (UAB) 0.91 0.86 0.80 0.86 Small partnership 0.07 0.13 0.19 0.12 Vilnius county 0.46 0.45 0.45 0.45 Kaunas county 0.20 0.20 0.20 0.20 Klaipeda county 0.11 0.11 0.11 0.11 Firms’ outcomes Gross Revenue (1,000s) 897.97 1011.04 995.97 966.63 (39.53) (44.36) (44.85) (43.47) Number employees 11.40 12.10 62.43 141.32 (2.00) (2.00) (2.00) (2.00) Taxable profit (1,000s) 91.37 88.80 89.15 87.15 (8.61) (7.02) (5.31) (6.29) Tax liability (1,000s) 12.84 12.50 12.53 12.25 (0.68) (0.48) (0.31) (0.44) Pre-tax profit margin (%) 2.84 4.11 6.41 4.32 (1.65) (1.85) (3.20) (1.94) N 88,696 98,543 104,711 579,276 Note: This table provides descriptive statistics of the corporate income tax data, for three separate years in columns 1 - 3 and for the pooled sample for 2015 - 2018 in column 4. The first panel presents the share of firms in each group, while the second panel presents averages of each variable woth medians in parentheses. 4 Bunching responses to the change in average tax rate We start this section documenting that, consistent with the incentives discussed above, firms behave strategically and systematically report turnover below the €300,000 threshold in order to avoid facing a higher 13 tax rate.7 These bunching responses can be easily detected through visual inspec- tion in Figure 2, where we present the distribution of declared turnover for all firms with revenue between €280,000 and €300,000 in the period 2015-2020.8 Similar to most countries, we observe a large amount of very small firms and a limited num- ber of firms with large turnover—reflected in the downward slope of the density presented. Precisely around the €300,000 threshold, nonetheless, we see an unusual spike: many firms declare turnover exactly below that level. We can also see a "missing" mass of firms immediately above the threshold: this response is precisely what we should expect given the incentive involved. Firms which would have re- ported turnover slightly above the threshold instead declare below it in order to benefit from the reduced rate. Bunching could be driven by two different underlying behaviors: real production decisions and/or misreporting. First, given the large increase in tax liability when crossing the threshold, firms have a real incentive not to increase their turnover: in the example discussed above, any profits gained from an extra €101 sales will be negligible next to the tax liability increase of €1,500 from facing the higher rate. So a store might decide not to sell another product to a customer or a provider might decide not to work an extra day to avoid generating revenue that will push them above the threshold. Second, firms might decide to still realize extra revenue that push their true turnover above €300,000, but not report that to the tax authority. Under this model, firms declaring turnover slightly below the threshold are likely misreporting part of their revenue. In reality, bunching is likely driven by a mix of the two behaviors and it is not always simple to disentangle and quantify the contribution of each.9 In the next section we quantify firms’ bunching responses and show how we can use them to learn about underlying firm behavior. 4.1 Quantifying firms’ responses In order to estimate how many firms were strategically locating below the threshold to avoid the increase in average tax rate, we follow several 7 In Appendix B we perform a series of robustness exercises meant to complement this analysis: (1) extend the time frame to 2011 (period prior to the adoption of the Euro); (2) define a balanced panel of firms; (3) document the relevance of the number of employees in the tax optimization behavior of firms; and (4) document how the bunching behavior varies across economic sectors. 8 In Figure 17, we present the same distribution but zoomed out to show revenues between €50,000 and €1,000,000. This alternate figure conveys the idea of exactly how skewed the firm-size distribution actually is. 9 Almunia and Lopez-Rodriguez (2018) and Lobel, Scot, and Zúniga (2021) use third-party reporting to show that, in Spain and Honduras, respectively, bunching behavior is partially driven by misreporting. 14 other studies in the literature of bunching responses (e.g. Saez (2010), Kleven (2016), Bachas and Soto (2021), among others). We provide more details about the methodology in Appendix A, but the intuition behind the exercise is that we consider what a "usual" distribution would look like in the absence of the incentive provided by differentiated tax rates. We do that by first excluding the region affected by the bunching behavior and "fitting" a curve to the observed distribution. We then extrapolate that prediction to the excluded region and obtain our estimates of bunching as the difference between the observed distribution of firms and the predicted values. Firms reduce their reported turnover by more than €30,000 to avoid facing the higher statutory tax rate. In Figure 3 below we present our estimates for the sample consisting of all firms filing between 2015 and 2020. The grey, smooth line in the figure is the fitted curve to the observed distribution of firms. Below the €300,000 threshold, we estimate an excess of approximately 1,150 firms—the shaded area in green. This mass is equivalent to almost five times the predicted number of firms we "should" be seeing immediately below the threshold, about 250 firms. If we observe this excess number of firms locating below the threshold, we should also observe a similar "missing" number of firms above, which we illustrate in the graph using the grey shaded area. The green and grey shades have the same area, since by our assumptions the firms that bunch below the threshold are coming from immediately above. In the graph we present our estimate of the marginal buncher : the firm with highest turnover that still decides to locate below the threshold. We estimate that number to be = €332,000: firms decrease their declared turnover by up to €32,000 to avoid the increase in tax liabilities. 15 Figure 3: Bunching below the €300,000 threshold Note: This figure presents empirical and counterfactual densities of declared turnover for a pooled sample of firms (2015 - 2020). The dashed line marks the €300,000 threshold while the dotted lines mark the lower and upper bounds of the bunching region. We present the excess mass below the notch (B), the excess mass as a share of the predicted mass in the bunching region (b), the upper bound obtained from the convergence method ( ) and the underlying number of taxpayers (N). Bins are €2,000 wide. We estimate that between 150 and 270 firms are bunching below the threshold each year. In Figure 4 we reproduce the same exercise separately by year. We highlight two features of that exercise. First, we note that the bunching behavior is increasing over time. For 2015, for example, we estimate an excess number of 140 firms below the threshold or four times the predicted mass of firms. That number increases to approximately 250 in 2018-2020 (six to seven times the predicted mass). This pattern is consistent with firms learning about the new incentives after the change in the notch location with the introduction of the Euro in 2015 and gradually adjusting their behavior over time. Second, bunching estimates are quite stable in the last two years of the period: we observe an excess mass between five and six times what is predicted and firms are decreasing their reported turnover by approximately €25,000-30,000 each year. 16 Figure 4: Bunching below the €300,000 threshold (a) 2015 (b) 2016 (c) 2017 (d) 2018 (e) 2019 (f) 2020 Note: This figure presents empirical and counterfactual densities of declared turnover for firms filing each year (2015-2020). The dashed line marks the €300,000 threshold while the dotted lines mark the lower and upper bounds of the bunching region. We present the excess mass below the notch (B), the excess mass as a share of the predicted mass in the bunching region (b), the upper bound obtained from the convergence method ( ) and the underlying number of taxpayers (N). Bins are €2,000 wide. The bunchers are high-profit firms, that gain more from the lower 17 tax rate. Not all firms face the same incentives to locate below the turnover threshold where the statutory tax rate increases. Consider for example firms facing losses. For them, paying 5% or 15% of zero taxable income is the same, so we should not expect them to respond by declaring lower revenue. Now consider two firms with the same turnover of slightly more than €300,000, one with a 1% profit margin and another with 10% margin. The low-profit firm would owe 15%*€3,000 = €450 in taxes staying above the threshold and 5%*€3,000 = €150 if they report below, therefore saving €300. The high-profit firm would owe €4,500 above and €1,500 below, saving €3,000 by bunching. The incentives to bunch are much larger for high-profit firms, so we should expect to see more of them bunching. This is precisely what we document on Figure 5: firms slightly below the threshold report a much higher profit margin, of up to 15% on average, when compared to firms with lower or higher revenue. This is a purely composition effect: bunchers are high-profit firms, so we observe a much higher average profit margin where they locate. Figure 5: Declared profit margin around €300,000 Note: This plot shows the average profit margin (and 95% CI of the average) for firms with different levels of declared turnover for a pooled sample of firms (2015-2020). Profit margins are winsorized at 1st and 99th percentiles. Bins are €2,000 wide. Firms strategically avoiding the higher tax rate translates into a loss of revenue of approximately €6 million for the tax administration between 2015-2020. We use the following estimates to provide a back of the envelope calculation of losses directly attributable to the bunching behavior: (i) we have 18 shown that approximately 1,200 firms bunch; (ii) by our assumptions the firms that bunch below the threshold are coming from immediately above, so most bunchers would otherwise have an average revenue of approximately €310,000; and (iii) based on the previous estimates on profits, we consider that bunching firms would have reported an average profit margin of 15%. If these firms had not optimized their behavior to face a lower tax rate, their average profit would have been 15%*310,000 = 46,500 and, applying a 15% tax rate we get €6,975 liability for each firm. After optimizing their behavior (keeping level of profits constant), the same firms face a 5% rate and pay €2,325 on average. Therefore in aggregate the tax revenue loss for the 2015-2020 period is approximately (€6,975-€2,325)*1,200 = €5,580,000. 5 Alternative response: changes in reported profit margins An additional margin of response to higher tax rates is the reporting of profits. In the previous section we documented one dimension of taxpayers’ response to the tax incentives generated by the 5% and 15% tax rate regimes: firms that would have reported turnover above the €300,000 threshold instead bunch below that level to avoid the increase in tax liability. In this section we document a second behavioral response to the higher tax rate regime: firms changing their reported taxable income. Higher tax rates on profits encourage firms to overreport their true costs. Firms whose turnover is significantly above the €300,000 threshold will most likely not bunch below that level: either by decreasing their sales or misreporting their turnover, it is probably too costly for them to declare much lower turnover. Nonetheless, the higher tax rate on their income creates another incentive. Under a 5% income tax rate, increasing the reported costs (deductions) by €1 leads to a €0.05 decrease in tax liability. Under the 15% tax rate regime, by contrast, the same increase in reported costs will reduce tax liability by €0.15. Since the incentive of firms to report taxable income changes so starkly, it is a natural question to investigate whether firms also change their reported profit margins. In Figure 6, we document that seems to be the case: firms located slightly above the €300,000 threshold (liable for a 15% tax rate) declare substantially lower profit margins when compared to those facing the 5% tax regime. On the left-hand side of the figure, we see that firms declaring turnover below €280,000 (those below the region where we observe bunching) declare pre-tax profit margins on the range of 5% to 7%. Absent the change in tax policy, there is no 19 reason to believe that firms that are slightly larger, with turnover around €350,000 for example, should have systematically different pre-tax profit margins. But that seems to be the case in the data: firms with turnover above €320,000 declare on average profit margins close to 2%-3%—three to four percentage points lower than those facing the reduced tax rate regime. This behavior is consistent with what has been seen in other contexts such as Costa Rica (Bachas and Soto 2021) where firms face similar incentives. That stark change in declared pre-tax profit margins around the threshold is consistent with a model where firms can over-report part of their costs to reduce their tax liability: when the incentives to do so become stronger due to an increase in tax rates, firms respond accordingly by increasing over-reporting. Figure 6: Declared profit margin around €300,000 Note: This plot shows the average profit margin (and 95% CI of the average) for firms with different levels of declared turnover for a pooled sample of firms (2015-2020). Profit margins are winsorized at 1st and 99th percentiles. Bins are €2,000 wide. In Table 2, we estimate a regression model that allows us to quantify exactly how much firms reduce their reported profit margins when facing a higher tax rate. Specifically, our regression has profit margin as dependent variable and evaluate whether, controlling for firm and year fixed-effects, as well as a linear trend of turnover, firms facing the 15% tax rate regime report lower margins. The answer is a compelling yes: we estimate that, compared to firms facing the 5% regime, they declare 5 percentage points lower profit margins—a similar result to what we observe in Figure 6. 20 Table 2: Profit margin as a function of turnover (1) (2) Profit margin Profit margin Turnover (1,000 euro) 0.000140*** 0.000481*** (0.0000177) (0.0000154) 15% rate bracket -0.0421*** -0.0502*** (0.00445) (0.00337) Turnover (320,340) 0.00583* 0.0219*** (0.00349) (0.00250) Turnover (300,320) -0.00319 0.0250*** (0.00342) (0.00268) Turnover (260,280) 0.00525 -0.00332 (0.00324) (0.00220) Turnover (280,300) 0.0440*** -0.00459* (0.00387) (0.00247) Constant 0.0279*** -0.0489*** (0.00338) (0.00295) Observations 111,419 111,419 R-Squared 0.007 0.728 Firm & Year FE? No Yes Note: This table reports a regression with profit margin as dependent variable. The coefficient on "15% rate bracket" indicates firms that report turnover above the €300,000 threshold. Profit margin is winsorized at 5th and 95th percentiles. Sample is restricted to firms filing in the 2015-2020 period and with turnover in the range €100,000-€500,000. Robust standard errors in parenthesis. 6 Recovering elasticities using bunching behavior When considering how firms change their behavior in response to tax policy, one key parameter to be considered are elasticities. The elasticity of (reported) turnover, for example, is by how much (in percentage terms) firms change their reported turnover when facing a 1% increase in the tax rate. An extensive literature has shown how one can recover the reported profit elasticity of firms by using their responses to tax notches such as the one we study in Lithuania. We provide detailed explanation of the methodology in Appendix A. The intuition for these exercises follows directly from the previous analysis: when firms face an increase in their turnover beyond the €300,000 threshold, they face a higher tax rate 21 and respond, as we have shown, by changing both their reported revenue (bunching) and their reported costs (driving a change in profit margins). We explore these behavioral responses and obtain their implied elasticities. When facing a 1% increase in the profit tax rate, firms around the notch reduce their reported profits by 7% to 10%. We summarize our findings in Table 3, separately for each year and for all years (2015-2020) pooled together. This is our preferred specification since it allows us to have a larger sample. For that period, as presented above, we estimate that firms reduce their turnover by up to €32,000 to locate below the threshold and enjoy lower tax rates. That is the key quantity that allows us to recover the elasticity of turnover, that we estimate to be 0.35—a 1% increase in turnover taxes decrease reported turnover by 0.35%.10 To estimate the elasticity of costs, we exploit the fact that firms below and above the €300,000 threshold report systematically different costs, which explains the changes in levels of profit margins we documented in the previous section. In Figure 7, we show that costs change linearly with reported revenue—but they "jump" around the threshold where the tax rate increases from 5% to 15%, allowing us to estimate the elasticity of costs. Consistent with findings from other settings (Bachas and Soto 2021), we find that costs are more elastic than revenues: firms change their reported costs by 1.3% when faced with a 1% increase in profit taxes. Putting these two results together, we can arrive at the total profit elasticity of 7.4 for the pooled years. We note that across years there is quite a substantial variation in our estimated elasticities, particularly for 2017 and 2018, when we estimate extremely high elasticities. For the last two years in our sample and the pooled period our estimates are more stable at 7% to 10%. 10 While there are no explicitly turnover taxes for corporations in Lithuania, the increase in average profit taxes above the threshold is equivalent to a turnover tax: by increasing their turnover, firms face a larger tax liability. 22 Table 3: Elasticities (1) (2) (3) (4) (5) Year Yu Cost (thousand) Cost Elasticity Revenue Elasticity Profit Elasticity 2015 327.00 291.32 -0.60 0.52 9.69 2016 313.00 233.26 -1.20 0.04 3.65 2017 337.00 297.30 -1.42 1.02 19.29 2018 360.00 300.73 -0.12 1.95 12.43 2019 327.00 247.36 -2.22 0.17 7.62 2020 324.00 260.61 -2.19 0.18 9.92 Pooled 332.00 269.56 -1.29 0.35 7.41 Note: This table reports the main results in terms of behavioral responses to the tax change at the threshold. Column (1) shows the revenue bin of the marginal buncher which is endogenously determined by the iteration procedure explained in Appendix A.1. Column (2) shows the 10 percentile of the cost distribution for the firms in . Column (3), (4) and (5) report the cost, revenue and profit elasticity, respectively. We report these numbers for each year since the Euro was introduced in Lithuania and for all years pooled together, which is our preferred specification. Figure 7: Average Costs 400 350 Average Costs 300 250 200 200 250 300 350 400 Revenues (in thousands) Note: This plot shows the average reported cost for each revenue bin around the threshold. The gray dashed lines correspond to and respectively. We use the bins the left of to estimate a linear fit of the cost-revenue relation below the threshold. We do the same above the threshold. This gives us the jump in cost around the threshold. 23 7 Dynamic incentives: does the notch affect firm growth? A firms’ response to the tax notch can be static or dynamic. The previous sections provided evidence that firms respond strategically to the threshold where tax rates change and locate below the notch. However, that is a static response : firms that, in any given year, would have located above the threshold in the absence of tax policy, decide to locate below. A different question is whether the existence of the notch affects the growth path of smaller firms. On the one hand, there exists broad evidence in the economics literature that size-dependent policies (not only tax policy) can affect the growth of firms (Bachas, Fattal Jaef, and Jensen 2019): providing strong benefits for small firms might discourage them from becoming larger. On the other hand, it is an empirical question whether any specific size- dependent policy is enough of a deterrent for firms’ growth. It is not easy to answer whether the notch generated by tax policy in Lithuania hampers firm growth. To start, it is not obvious to determine which firms are affected by the existence of the threshold. It is expected that firms with revenue slightly below €300,000 are aware of the change in incentives if they grow past that limit. But it is possible that all firms with turnover below €300,000 take those incentives into consideration and therefore adjust their behavior considering the future possibility of growing "too much" and starting to pay taxes. In this section we provide evidence that in the short run—that is, from one year to the next—some firms report lower growth when facing the immediate possibility of crossing the threshold. For longer periods the evidence is less compelling and, therefore, we avoid making any claims about the effects of this specific notch on the long-term growth pattern of SMEs. For the short run response, we find that firms immediately below the threshold in one year will grow systematically less than other firms in the economy to avoid paying higher taxes. In Figure 8 we present how the growth pattern of firms differ across different levels of turnover for the period 2016-2020. Notice that the pattern of growth is very similar at most turnover levels: when we observe a firm in year t-1 declaring turnover between €100,000 and €500,000, the median turnover growth expected in the following year is between 1% - 2%. There is one very stark exception in that figure: if we observe a firm declaring turnover between €280,000 and €300,000, immediately below the notch, the median turnover growth observed in the following year is zero - more than half of firms in that turnover bin will declare zero or negative growth the following year. 24 An alternative way to visualize that change in behavior is looking at the distribution of firm growth in each bin, presented in Figure 9. While the density of growth is very similar for bins farther from the threshold, those immediately below (€280,000 - 300,000) have disproportionate growth below or at zero - compared to other firms, there is a clear lack of firms with positive but small growth. Given the incentives for firms not to grow above the threshold, this is compelling evidence that, at least in the short run, firms avoid growing (at least on their declared turnover to the tax authority) when approaching the threshold. Figure 8: Revenue dynamics - growth rate around €300,000 Note: This figure shows the median of yearly turnover growth for firms in the panel (2016- 2020), for different bins of turnover in the initial year. Bins are €20,000 wide. 25 Figure 9: Revenue dynamics - density of growth rate around €300,000 Note: This figure shows the density of yearly turnover growth for firms in the panel (2016- 2020), for different bins of turnover in the initial year. Bins are €20,000 wide Evaluating the medium- to long-term growth effects of the tax notch is a more challenging task, precisely because it is not clear which firms will eventually be affected. We provide evidence on the medium-term growth pattern for two specific groups of firms: those that in 2015 declared turnover slightly below the €300,000 and those that in the same year declared turnover slightly above. The idea is that those below are likely to consider the incentives provided by the threshold and grow less, while those above will not be affected by it. Over a five year period, firms starting slightly below the threshold grow less than those above, but the effect is concentrated in the initial years. In Figure 10 we present the median growth rate for these groups of firms, for each year between 2016 and 2020. For the 2016, we observe the same pattern as documented for short-term growth above: while firms with turnover above €300,000 declare a medium growth rate of 2%, those below declare virtually zero growth. The difference in growth rates is also observed in the second year, when larger firms register growth rate of 5% vs. 2% for those below the threshold. Nonetheless, after the second year the picture becomes less clear: firms below the threshold grow slightly more in 2018, then less in 2019 and negative but smaller in magnitude growth rates in 2020 (where most firms declare negative growth rates, likely due to the shock of Covid-19 pandemic). These differences in the yearly growth in the first two periods are sufficient to 26 generate some medium-term effects. When we observe the median growth in the full period 2015-2020, the median growth of those starting below the threshold is almost zero while those above the threshold grow by about 1%. The picture is similar if we exclude 2020, when firms were severely affected by the Covid-19 crisis. While the median growth is much higher for the two groups, the gap between them is similar: firms starting above €300,000 grew by 3% while those below grew less than 2%. Figure 10: Revenue dynamics - growth rates in the medium-run .06 Median Revenue Growth Rate -.02 0 .02 -.04 .04 16 17 18 19 20 ) ) 20 19 20 20 20 20 20 20 20 5- 5- 01 01 (2 (2 [350k; 450k] [270k; 300k] Note: This figure reports median yearly revenue growth in each year between 2016 and 2020, for firms reporting revenue between €350,000-450,000 in 2015 and those reporting €270,000- 300,000. The ranges of initial revenue are different so that both groups have the same amount of firms. Numbers for the pooled periods of 2015-2019 and 2015-2020 are for median annualized growth rates. An alternative way to assess the dynamic effects of the tax notch is to ask whether firms below the threshold are likely to get "stuck" in the €300,000 level. In Figure 11 we compute the share of firms declaring turnover below the threshold in one year that end up above the threshold in the following year. For firms with revenue between €260,000-300,000 in one year, almost 40% will declare turnover above that level in the next year. This tells us that several firms will not be deterred by the increase in tax rate and will grow into the higher tax-rate bracket. But this cannot tell us what we should expect growth to be, in the absence of the tax notch. To provide evidence on that question, in Figure 12 we show, for different horizons, what is the share of firms declaring growth of €20,000 or more. The orange line shows that, on a one-year period, those declaring revenue 27 in the interval €280,000-300,000 are much less likely to declare growth above that level than other firms with lower or higher revenues—meaning that they are more likely to stay below the threshold. On a five-year period (2015-2020), however, that effect disappears: firms starting below the threshold are no less likely to have grown enough to surpass the threshold providing beneficial tax rate. This is additional evidence that, while the notch seems to have very meaningful effects in the short run, in the medium-run it seems to be less of a stark deterrent to the growth of all small firms. Figure 11: Crossing the threshold .4 Proportion going above the threshold .1 .2 0 .3 -5 -4 -3 -2 -1 Distance to Threshold (1 = 40k from the threshold) Note: This plot shows, for each revenue bin, what proportion of those who were below 300,000 euros in revenues in t, end up above such threshold in t+1. 28 Figure 12: Proportion of firms reporting turnover growth above €20,000 .6 5-year period Proportion growing 20k or more .4 .5 1-year period .3 .2 100 200 300 400 500 Turnover t0 Note: This figure takes turnover intervals of size €20,000 and shows the share of firms within each interval declaring turnover growth of €20,000 or more. The orange line shows the propor- tion for a one-year interval, while the blue line calculates the proportion for a five-year interval. 8 Conclusion This report studies how legal entities in Lithuania respond to the exis- tence of differential tax rates based on firm size by leveraging administra- tive microdata on the universe of limited liability corporate taxpayers. Corporations in Lithuania face an increase in the average tax rate when they declare turnover above €300,000—this notch in the tax schedule creates a sudden increase in average tax rates from 5% to 15%. The use of preferential tax treatment tar- geted towards entrepreneurs and small and medium enterprises as the one described above is usually justified on the grounds of providing support to firms in their early stages. However, this type of policy may also provide the wrong incentives for firms who may take advantage of the special treatment by remaining small in order to pay lower taxes. We document a clear response of firms to this policy in the short run, while this effect weakens as we extend the time horizon. In the short run, firms respond to the incentives that are present in the tax schedule by reporting turnover such that they remain below the threshold and by overreporting costs. Firms reduce their reported turnover by up to €32,000 to avoid facing the higher statutory tax rate. This bunching response could be driven by two different underlying behaviors: real production 29 decisions and/or misreporting. However, an additional finding suggests it is mostly misreporting. We document that higher tax rates on profits encourage firms to overreport their true costs and therefore small firms declare higher pre-tax profit margins than large firms. Absent the change in tax policy, there is no reason to believe that firms that are slightly larger should have systematically different profit margins. Therefore, this discontinuity in firms’ costs at the revenue threshold is more compatible with misreporting than real reduction in economic activity. Finally, by quantifying the response on both reported turnover (bunching) and reported costs (change in profit margins) we are able recover the reported profit elasticity of firms: when facing a 1% increase in the profit tax rate, firms close to the notch reduce their reported profits by 7-10%. The evidence of firms responding in the medium- to long-run is less clear. The main difficulty faced when trying to determine how the differential tax rates affect the long term growth of firms is that it is not obvious which firms are affected by the existence of the threshold (even firms far from it may adjust their behavior in response) which complicates the definition of a group that can serve as a benchmark. We provide insights that firms facing the immediate possibility of crossing the threshold react by reporting lower growth in turnover. This effect is clear for year-on-year growth calculations, but dissipates as we consider longer time spans. 30 References Abbring, Jaap H., and Jeffrey Campbell. 2005. A Firm’s First Year. Tinbergen Institute Discussion Paper 05-046/3. Tinbergen Institute, May. https://econp apers.repec.org/paper/tinwpaper/20050046.htm. Akcigit, Ufuk, John Grigsby, Tom Nicholas, and Stefanie Stantcheva. 2018. Taxation and Innovation in the 20th Century. Working Paper, September. https://www. nber.org/papers/w24982. Almunia, Miguel, and David Lopez-Rodriguez. 2018. “Under the Radar: The Effects of Monitoring Firms on Tax Compliance” [in en]. American Economic Journal: Economic Policy 10, no. 1 (February): 1–38. Bachas, Pierre, Roberto N. Fattal Jaef, and Anders Jensen. 2019. “Size-dependent tax enforcement and compliance: Global evidence and aggregate implications” [in en]. Journal of Development Economics 140 (September): 203–222. https: //www.sciencedirect.com/science/article/pii/S0304387818308447. Bachas, Pierre, and Mauricio Soto. 2021. “Corporate Taxation under Weak Enforce- ment” [in en]. American Economic Journal: Economic Policy Forthcoming. Bilicka, Katarzyna Anna. 2020. Are Financing Constraints Binding for Invest- ment? Evidence From a Natural Experiment [in en]. SSRN Scholarly Paper ID 3682287. Rochester, NY: Social Science Research Network, May. https : / / papers.ssrn.com/abstract=3682287. Chetty, Raj, Alex Bell, Xavier Jaravel, Neviana Petkova, and John Van Reenen. 2019. Do tax cuts produce more Einsteins? The impact of financial incentives vs. exposure to innovation on the supply of inventors [in eng]. Monograph. ISSN: 2042-2695 Issue: 1597 Num Pages: 29 Number: 1597 Place: London, UK Publisher: Centre for Economic Performance, LSE, January. http://cep.lse.ac. uk/_new/publications/series.asp?prog=CEP. Chetty, Raj, and Emmanuel Saez. 2005. “Dividend Taxes and Corporate Behavior: Evidence from the 2003 Dividend Tax Cut.” Publisher: Oxford University Press, The Quarterly Journal of Economics 120 (3): 791–833. https://www.jstor.org/ stable/25098756. 31 Cummins, Jason G., Kevin Hassett, and R. Glenn Hubbard. 1994. “A Reconsid- eration of Investment Behavior Using Tax Reforms as Natural Experiments.” Publisher: Economic Studies Program, The Brookings Institution, Brookings Papers on Economic Activity 25 (2): 1–74. https : / / econpapers . repec . org / article/binbpeajo/v_3a25_3ay_3a1994_3ai_3a1994-2_3ap_3a1-74.htm. Denes, Matthew, Xinxin Wang, and Ting Xu. 2019. “Financing Entrepreneurship: Tax Incentives for Early-Stage Investors” [in en]. SSRN Electronic Journal, https://www.ssrn.com/abstract=3454633. Djankov, Simeon, Tim Ganser, Caralee McLiesh, Rita Ramalho, and Andrei Shleifer. 2010. “The Effect of Corporate Taxes on Investment and Entrepreneurship” [in en]. American Economic Journal: Macroeconomics 2, no. 3 (July): 31–64. https://www.aeaweb.org/articles?id=10.1257/mac.2.3.31. Gechert, Sebastian, and Philipp Heimberger. 2022. “Do corporate tax cuts boost economic growth?” [In en]. European Economic Review 147 (August): 104157. https://www.sciencedirect.com/science/article/pii/S0014292122000885. Gordon, Roger, and Wei Li. 2009. “Tax structures in developing countries: Many puzzles and a possible explanation.” Publisher: Elsevier, Journal of Public Eco- nomics 93 (7-8): 855–866. Harju, Jarkko, Aliisa Koivisto, and Tuomas Matikka. 2022. “The effects of corpo- rate taxes on small firms” [in en]. Journal of Public Economics 212 (August): 104704. https://www.sciencedirect.com/science/article/pii/S00472727220010 62. Hurst, Erik, and Benjamin Wild Pugsley. 2011. What Do Small Businesses Do? Working Paper, May. https://www.nber.org/papers/w17041. Ito, Koichiro, and James Sallee. 2014. The Economics of Attribute-Based Regulation: Theory and Evidence from Fuel-Economy Standards [in en]. SSRN Scholarly Paper ID 2499342. Rochester, NY: Social Science Research Network, Septem- ber. Jacob, Martin. 2021. “Dividend taxes, employment, and firm productivity” [in en]. Journal of Corporate Finance 69 (August): 102040. https://www.sciencedirect. com/science/article/pii/S0929119921001620. Kalanta, Marius. 2021. Special Corporate Income Tax Rate for Micro-Enterprises in Lithuania: Productive or Unproductive Incentive? Technical report. 32 Kleven, Henrik J. 2016. “Bunching.” Annual Review of Economics 8 (1): 435–464. Kleven, Henrik J., and Mazhar Waseem. 2013. “Using Notches to Uncover Optimiza- tion Frictions and Structural Elasticities: Theory and Evidence from Pakistan” [in en]. The Quarterly Journal of Economics 128, no. 2 (May): 669–723. Lobel, Felipe, Thiago Scot, and Pedro Zúniga. 2021. “Corporate Taxation and Eva- sion Responses: Evidence from a Minimum Tax in Honduras” [in en], 89. Londoño-Vélez, Juliana, and Javier Ávila-Mahecha. 2019. “Can Wealth Taxation Work in Developing Countries? Quasi-Experimental Evidence from Colombia.” Working Paper, Job Market paper. Manoli, Day, and Andrea Weber. 2016. “Nonparametric Evidence on the Effects of Financial Incentives on Retirement Decisions” [in en]. American Economic Journal: Economic Policy 8, no. 4 (November): 160–182. https://www.aeaweb. org/articles?id=10.1257/pol.20140209. Matray, Adrien, and Charles Boissel. 2020. Higher Dividend Taxes, No Problem! Evidence from Taxing Entrepreneurs in France [in en]. Technical report 276. Publication Title: Working Papers. Princeton University, Department of Eco- nomics, Center for Economic Policy Studies., September. https://ideas.repec. org/p/pri/cepsud/276.html. Moon, Terry S. 2022. “Capital Gains Taxes and Real Corporate Investment: Ev- idence from Korea” [in en]. American Economic Review 112, no. 8 (August): 2669–2700. https://www.aeaweb.org/articles?id=10.1257/aer.20201272. Ohrn, Eric. 2018. “The Effect of Corporate Taxation on Investment and Financial Policy: Evidence from the DPAD” [in en]. American Economic Journal: Eco- nomic Policy 10, no. 2 (May): 272–301. https://www.aeaweb.org/articles?id= 10.1257/pol.20150378. . 2019. “The effect of tax incentives on U.S. manufacturing: Evidence from state accelerated depreciation policies” [in en]. Journal of Public Economics 180 (December): 104084. https://www.sciencedirect.com/science/article/pii/ S0047272719301458. Ohrn, Eric, E. Mark Curtis, Daniel G. Garrett, Kevin A. Roberts, and Juan Carlos Suarez Serrato. 2022. “Capital Investment and Labor Demand” [in en], 125. 33 Ravšelj, Dejan, Polonca Kovač, and Aleksander Aristovnik. 2019. “Tax-Related Bur- den on SMEs in the European Union: The Case of Slovenia” [in en]. Mediter- ranean Journal of Social Sciences 10, no. 2 (March): 69–79. http://archive. sciendo.com/MJSS/mjss.2019.10.issue-2/mjss-2019-0024/mjss-2019-0024.pdf. Saez, Emmanuel. 2010. “Do Taxpayers Bunch at Kink Points?” [In en]. American Economic Journal: Economic Policy 2, no. 3 (August): 180–212. Saez, Emmanuel, Benjamin Schoefer, and David Seim. 2019. “Payroll Taxes, Firm Behavior, and Rent Sharing: Evidence from a Young Workers’ Tax Cut in Sweden” [in en]. American Economic Review 109, no. 5 (May): 1717–1763. https://www.aeaweb.org/articles?id=10.1257/aer.20171937. Sallee, James M., and Joel Slemrod. 2012. “Car notches: Strategic automaker re- sponses to fuel economy policy” [in en]. Journal of Public Economics 96, nos. 11-12 (December): 981–999. Shane, Scott. 2009. “Why encouraging more people to become entrepreneurs is bad public policy” [in en]. Small Business Economics 33, no. 2 (August): 141–149. https://doi.org/10.1007/s11187-009-9215-5. Sterk, Vincent, Petr Sedláček, and Benjamin Pugsley. 2021. “The Nature of Firm Growth” [in en]. American Economic Review 111, no. 2 (February): 547–579. https://www.aeaweb.org/articles?id=10.1257/aer.20190748. Wang, Yao. 2016. “What are the biggest obstacles to growth of SMEs in developing countries? – An empirical evidence from an enterprise survey” [in en]. Borsa Istanbul Review 16, no. 3 (September): 167–176. https://www.sciencedirect. com/science/article/pii/S2214845016300539. Yagan, Danny. 2015. “Capital Tax Reform and the Real Economy: The Effects of the 2003 Dividend Tax Cut” [in en]. American Economic Review 105, no. 12 (December): 3531–3563. https://www.aeaweb.org/articles?id=10.1257/aer. 20130098. Zwick, Eric, and James Mahon. 2017. “Tax Policy and Heterogeneous Investment Behavior” [in en]. American Economic Review 107, no. 1 (January): 217–248. https://www.aeaweb.org/articles?id=10.1257/aer.20140855. 34 Appendix A Bunching Estimators This section explains in detail how we relate bunching behavior and changes in costs with the elasticities estimated in Table 3. The intuition is that, under some assumptions on the underlying behavior of firms, we can recover "structural" pa- rameters that govern their decisions from behavioral responses to taxes changes. Different from the classic literature in bunching estimators started by Saez (2010), this setting presents two key challenges. First, the variable which determines the tax rate (turnover) is not the tax base (profits or taxable income). The amount of turnover (above or below €300,000) determines the fraction of profits that should be paid in taxes. This implies that firms have more than one option to respond to tax incentives. They can report less turnover to keep a lower tax rate or over-report costs to reduce profits. This is why we present not only revenue elasticities, but also cost and profit elasticities in our results. Second, and as a consequences of the previous point, the revenue threshold introduces a notch rather than a kink. This means that it is not just a change in the marginal tax rate, but the increase in the tax rate affects the whole tax base. For that reason, we use the formulas developed in Kleven and Waseem (2013) rather that simple elasticities used by Saez (2010). The methodology we are implementing is also similar to the one used by Bachas and Soto (2021). The profit elasticity ( ) is simply the percentage change in profits when the tax rate changes in 1%. Since profits is turnover minus costs (deductions), we can re-write this expression as a function of the elasticity of turnover ( ) and costs ( ): Δ 1 − ,1− · − ,1− · ,1− = = (1) Δ A.1 Revenue Elasticity To calculate the revenue elasticty (,1− ) we use the patterns of excess and missing mass (exemplified in Figure 3) to estimate the change in revenue Δ * triggered by the tax change for the marginal bunching firm. The idea is to transform the average tax variation at a notch into the equivalent marginal tax variation that would occur at a kink, which changes the marginal tax rate from to * . On top of that, we need to adapt the notch formula to take into account that our tax base is profit and not revenues. The formula, the same used by Bachas and Soto (2021), is as follows: 35 1 (Δ * / )2 ,1− ≈ · (2) 2( − * )/ * + Δ * / /(1 − 0 ) If we ignore extensive margin effects (firms leaving the sample), then the excess mass below the threshold must equal the missing mass above it. Therefore, in order to find the marginal buncher, we first determine in which revenue bin the excess bunching begins. We then estimate a fifth-order polynomial regression, in which we try to "fit" the distribution of firms according to their turnover, but ignoring the region affected by the bunching. The regression takes the form: 5 ∑︁ ∑︁ = ( ) + 1( = ) + (3) =0 = Where is the number of firms in revenue bin and we use as the revenue midpoint of interval . [ , ] is the excluded region and ’s are dummy shifters for the excluded regions. Our prediction for what the distribution should be, in the absence of bunching, is the fitted model without the dummies: ^ = ∑︀5 ^ =0 ( ) . Our estimate of excess mass below the threshold ( ^ ), therefore, is the difference between the observed number of firms per bin and the predicted number: ^ = − ^ . Note that affects the polynomial fit (since it excludes more or less revenue bins) and it is also affected by it (since it changes the excess mass). Therefore, it is determined by iterating equation (3) until it finds the first revenue bin to the right of the threshold that satisfies the equality between excess and missing mass. With this method we recover the marginal buncher = * which allows us to calculate Δ * . The last unknown in equation (2) is * , which are the total cost of the marginal buncher. We set this number to be the 10th percentile of the cost distribuition within the marginal buncher revenue bin.11 A.2 Cost Elasticity As we mentioned before, firms can reduce their tax base also by increasing reported costs. In Figure 7, we see a clear increase in reported costs to the right of the threshold. Importantly, some firms have selected into the revenue range around the threshold, as a function of their costs. We know that this selection occurs between [ , ]. Therefore, we use the donut-hole discontinuity method to measure the discontinuity in cost at the threshold adding dummy variables for the bins 11 From theory we know that the marginal buncher is the firm with the lowest costs within its revenue bin. 36 within the selection interval. ∑︁ = + · 1( > 0) + 1 · 1( > 0) + 1( = ) + (4) = In this specification we use = − , which is the revenue distance to the threshold (so we are centering the revenue variable in the threshold). are the dummy shifters for the bins in the excluded region. Moreover, 1 recovers the slope below the threshold and 1 + 2 the slope above. Finally, is the jump in cost for crossing the threshold, and / is the percentage increase in costs. One more thing must be taken into account. The cost elasticity aims to recover the change in cost due to the tax change at the threshold holding revenue response constant. In the previous specification, we cannot rule out intensive-margin response of revenue. However, we can use the revenue elasticity explained in Appendix A.1 and adjust revenue to control for intensive margin responses: = if ≤ (5) = + ^,1− · · 1− if > Table (4) provides the results for the adjusted and non-adjusted regressions and ^ / ^,1− = Δ we can recover the cost elasticity from ^ /(1− ) . 37 Table 4: Donut-Hole Discontinuity Regression (1) (2) Cost Cost (Adjusted) Jump in cost 36.63*** 28.07*** (4.902) (5.462) Slope below threshold _1 0.603*** 0.603*** (0.0396) (0.0396) Slope change past threshold _2 0.203*** 0.174** (0.0733) (0.0714) Cost at threshold 207.3*** 207.3*** (2.408) (2.408) Observations 200 200 R-Squared 0.980 0.980 Note: This table reports the estimated coefficients from equation (4). Column (1) present the specification without revenue adjustment, which means it incorporate intensive-margin responses in revenues. Column (2) controls for these intensive margin responses by adjusting revenues using equation (5) in Appendix A.2. B Robustness exercises In this section we provide several additional exercises that reinforce the findings discussed previously. Extended time frame. The preferential 5% tax rate for SMEs already existed before the introduction of the Euro in 2015: firms declaring revenue below 500,000 litas in 2011 and 1,000,000 litas in 2012 - 2014 also benefitted from the reduced rate. In Figure 13, we show that the same bunching behavior observed in 2015- 2020 was also present to some extent in the previous period. In 2011, when the threshold was 500,000 litas, we observe some amount of bunching at that level, and the bunching moves to 1,000,000 litas when the policy changes in 2012-2014. Immediately after the introduction of the Euro, the bunching again moves from the equivalent 1,000,000 litas to the new €300,000 threshold—reinforcing the message that the bunching behavior is indeed a response to the change in tax rates. 38 Figure 13: Bunching in the pre-Euro era Note: This plot shows the distribution of firms around the two threshold that existed before Euro was introduced. They were 500k litas in 2011 and 1M litas between 2012-2014. The exchange rate was fixed at 3.45 litas per Euro. Balanced panel of firms. In our main analyses, we use an unbalanced panel of firms - not all firms file every year, so our sample is changing every year. While there is no reason to believe that firms are exiting or entering precisely around the €300,000 threshold, we replicate our main finding of the bunching behavior for a balanced panel of firms that file every year in Figure 14. The result is qualita- tively similar: we observe the same behavior of an excess number of firms declaring turnover precisely below the level that separates the 5% and 15% tax rate regimes. 39 Figure 14: Bunching below the €300,000 threshold - Balanced panel Note: This figure presents the empirical density of turnover for a balanced panel of firms that file income taxes every year in the period 2015 - 2020. Panel. Bins are €2,000 wide. Number of employees and bunching behavior. While our focus through- out this report is on the turnover threshold that separates the two regimes, the number of employees is also a determinant of the tax rate faced by firms: the pref- erential 5% tax rate is only applied to firms that declare less than €300,000 in turnover and also declare less than 10 employees on average throughout the year. In Figure 15, we show that the level of employment matters for the bunching be- havior. The figure presents a heatmap, where lighter colors indicate bins (intervals of employment and turnover) where there are more taxpayers. First, note the clear bunching pattern around the €300,000 level: the bright colors to the left of the threshold immediately disappear to the right of it. But note also that this behav- ior only exists for firms with less than 10 employees: for those above, there seems to be no bunching in turnover, consistent with the fact that larger firms do not benefit from the 5% tax rate, regardless of their turnover level. There is also some evidence that firms bunch below the 10 employee level, with a slight concentration of firms below that level. Since the vast majority of firms are much smaller than that, however, it is harder to precisely estimate the importance of that behavior. 40 Figure 15: Bunching in Turnover vs. Employment Note: . Economic sectors. In Figure 16 we document that bunching behavior is widespread but not equal across economic sectors. We observe particularly large responses in construction, wholesale, real estate and other services. Potentially sur- prising, we actually see no bunching in retail, and restaurant and hotels—sectors that are often seen as at-risk for misreporting taxes. 41 Figure 16: Bunching below €300,000 threshold - by economic sector Note: This figure presents the empirical density of turnover for firms in the period 2015 - 2020. The dashed line marks the €45,000 threshold for VAT filing. Bins are €500 wide. 42 C Additional figures Figure 17: Empirical density of turnover (2015 - 2020) Figure 18: Between €50,000 and €1,000,000 Note: This figure presents the empirical density of turnover for firms in the period 2015-2020 for firms with revenue between €50,000 and €1,000,000. Bins are €2,000 wide. 43