14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License CYCLICAL TAX ENFORCEMENT JOSÉ M DURÁN-CABRÉ, ALEJANDRO ESTELLER-MORÉ and LUCA SALVADORI∗ We wonder whether tax enforcement varies along the economic cycle and aim at answering that question from a positive perspective by means of survey data for the Spanish case (1994–2015). According to a fiscal capacity argument, tax enforcement might be stronger in times of crisis (counter-cyclical), but if the tax administration prioritizes taxpayers’ welfare over public revenue, enforcement might be slacker (procyclical). We find tax enforcement is not immune to the state of the economy. In particular, it presents a prevailing counter-cyclical trend, but in presence of a severe economic crisis it turns out to be procyclical. (JEL D78, H12, H26, H83) I. INTRODUCTION of 3 percentage points) and even some increased top marginal personal income tax rates (e.g., 13 The economic downturn associated with the out of the 28 EU, with an average increase of global financial crisis caused a significant fall about 6 percentage points). But given the multi- in tax revenues in many countries. In advanced dimensional nature of tax systems, enforcement economies, fiscal deficit increased by 2.5% of is another parameter in the hands of the public gross domestic product (GDP) in 2008 and by sector—through its tax administration—to col- about 5% in 2009, provoking serious concern lect more revenues (Slemrod and Gillitzer 2014). about the need to substantially lower their deficits In other words, if tax administrators act as rev- to be able to control their debt-to-GDP ratios enue maximizers, they should increase enforce- (IMF 2010). In response, between 2008 and 2013, ment efforts; that is, they should have played a many countries augmented the value-added tax counter-cyclical role.11 (VAT) rates (e.g., 19 out of the 28 EU countries However, in a severe economic recession raised the general rate, with an average increase and with financial credit severely constrained, taxpayers’ aggregate welfare likely will be ∗ We are grateful to Umberto Galmarini, Amedeo Piolatto, declining. If tax administrations prioritize tax- Adam Spencer, two anonymous referees, and the Editor for payers’ welfare over revenue collection, they their helpful comments. We would also like to thank partici- pants at the seminar of the Department of Economics of the 1. Analyzing the finances of the southern European coun- University of Ferrara; XXVI Meeting on Public Economics tries, The Economist indicated “Now that these countries are (University of Oviedo), 5th TARC Workshop on Tax Admin- trying to get their finances in order, bringing down rates of tax istration (University of Exeter), and XXVIII SIEP Confer- evasion is a high priority” (August 12, 2010). Between 2010 ence (University of Lecce) for their useful comments. This and 2012, the annual Eurostat publication Taxation Trends in work was supported by the Instituto de Estudios Fiscales the European Union indicates that southern European coun- (Ministerio de Hacienda) and the RTI2018-095983-B-I00 tries (e.g., Greece, Italy, Portugal, Spain) but also other Euro- (MCIU/AEI/FEDER, UE). Durán-Cabré and Esteller-Moré pean governments (e.g., Belgium, Bulgaria, Norway) intro- also acknowledge funding from the project 2017SGR796 duced changes in the administration of taxes in order to fight (Generalitat de Catalunya). Salvadori also acknowledges against tax evasion and raise revenues. funding from the Economic and Social Research Council (grant number ES/S00713X/1). The usual disclaimer applies. Durán-Cabré: Associate Professor, Department of Eco- nomics, Universitat de Barcelona & IEB, 08034, ABBREVIATIONS Barcelona, Spain. E-mail jmduran@ub.edu AC: Autonomous Communities Esteller-Moré: Professor, Department of Economics, Univer- sitat de Barcelona & IEB, 08034, Barcelona, Spain. E- GDP: Gross Domestic Product mail aesteller@ub.edu FIS: Fully Interacted Specification Salvadori: Postdoctoral Research Fellow, University of GVA: Gross Value Added Exeter Business School—Tax Administration Research INE: National Institute of Statistics Centre (TARC) & IEB, Department of Economics, OLS: Ordinary Least Squares TARC, Streatham Court, Rennes Drive, EX4 4PU, UR: Unemployment Risk Exeter, UK. E-mail l.salvadori@exeter.ac.uk VAT: Value-Added Tax 1874 Economic Inquiry (ISSN 0095-2583) doi:10.1111/ecin.12902 Vol. 58, No. 4, October 2020, 1874–1893 Online Early publication May 11, 2020 © 2020 The Authors. Economic Inquiry published by Wiley Periodicals LLC on behalf of Western Economic Association International. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1875 should lessen enforcement and play a procyclical confounders that could vary over the cycle. We role. In other words, a procyclical tax agency identify two dimensions of the perceived enforce- will reduce the level of enforcement during a ment by individuals that can act in this way. time when credit-constrained taxpayers may Namely, these are individual risk perception and be tempted to use tax evasion as an alternative normative beliefs. We propose an identification source of finance, as they may perceive the risks strategy to deal with these two confounders. of tax evasion (penalties) much smaller than the According to our empirical results, perceived potential gains of avoiding bankruptcy (see e.g., tax enforcement presents a prevailing counter- Brondolo 2009). Regarding taxpayers’ behavior, cyclical trend, but becomes procyclical in pres- in a recent paper, Alm, Liu, and Zhang (2019) ence of severe economic downturns. This is the analyze whether the financial constraints faced main contribution of this paper. by a firm increase the extent of tax evasion The rest of the paper is organized as follows: and, working with a survey of firms from 27 Section II sets the main hypotheses concerning transitional countries, find evidence that more the impact of the cycle on tax enforcement; financially constrained firms are more likely to Section III presents the empirical strategy we be involved in tax evasion activities. Similarly, employ to test those hypotheses; Section IV Richardson, Taylor, and Lanis (2015), based presents the results of the empirical analysis, and on a sample of 203 publicly-listed Australian Section V concludes. An online appendix reports firms covering the 2006–2010 period, show the results of robustness analyses. that firms’ financial distress is significantly and positively associated with tax avoidance across several proxy measures of tax avoidance and II. TAX ENFORCEMENT ALONG THE ECONOMIC CYCLE financial distress. This idea was first analyzed by Andreoni Might tax enforcement be influenced by eco- (1992) for individuals in a theoretical model nomic cycles and particularly by a severe eco- in which the tax administration could act as nomic downturn? This is the main research ques- a last-resort lender (“the tax agency as a loan tion we pose. To counteract the negative macroe- shark”). When taxpayers face binding financial conomic effects of a downturn, tax policy should constraints, they may consider evading taxes to be counter-cyclical. Nonetheless, tax revenues smooth consumption along time. They would do also vary endogenously with changes due to fac- so even if evasion were not a fair gamble, that tors not under the control of policymakers, such is, the expected return from evasion was nega- as changes in income distribution or changes in tive. Only severely financial constrained taxpay- agents’ behavior over the economic cycle (Vegh ers would act like this, something very relevant and Vuletin 2015). at the aggregate level in times of crisis. The role Likewise, we also know the limits of fiscal pol- of tax evasion as a substitute for loans was also icy when the public debt-to-GDP ratio is high analyzed by Fishlow and Friedman (1994), in a (Ghosh et al. 2013), while some authors sim- paper where they focus on the public resort of ply cast doubts on the effectiveness of that pol- tax evasion in developing countries. They use a icy (Ramey and Zuvairy 2018), or suggest their theoretical model of intertemporal consumption impact—through the estimated value of the fis- that characterizes the behavior of taxpayers in a cal multipliers—is contingent on the state of the financially constrained economy and show that economy (Auerbach and Gorodnichenko 2012). negative shocks over current income raise eva- Being aware of those potential limitations, not sion. Agents use evasion to substitute for loans all fiscal instruments might be equally effective in economies where credit is not available. at stabilizing the economy (Wren-Lewis 2010). The aim of this paper is to analyze tax enforce- Under this context, tax enforcement might be a ment over the economic cycle. We do so by means peculiar instrument, which to our knowledge has of ordered response models applied to Spanish never been formally considered as a potential tool data extracted from repeated surveys and other for stabilization purposes. sources. In particular, we employ a measure of The most common approach sees tax admin- tax enforcement as it is perceived by individ- istration as a public agency whose aim is maxi- uals and try to identify what part of its vari- mizing tax revenues given a certain budget (e.g., ation along the economic cycle is due to the Shaw, Slemrod, and Whiting 2009; Slemrod and annual variation in the effort of the administra- Yitzhaki 1987, 2002). But, as we said before, tion. This entails isolating potential unobserved tax enforcement policies carried out by the tax 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1876 ECONOMIC INQUIRY administration might be linked to the economic especially where formal automatic stabilizers are cycle. Furthermore, recent empirical studies sug- weak. In particular, he cites two peculiar features: gest that political as well as budgetary variables (1) it is relative quick acting through it, as no play a role in determining tax administration’s legislative action is necessary and (2) it might enforcement efforts (see, e.g., Baretti, Huber, and be targeted to those more in need, if changes in Lichtblau 2002; Bönke, Jochimsen, and Schröder the level of tax evasion indicates changing liquid- 2017; Esteller-Moré 2005, 2011; Young, Reksu- ity constraints among taxpayers. In other words, lak, and Shughart II 2001). tax administration could be an expeditious tool A strand of literature underlines the institu- to deal with asymmetric shocks within a country. tional capacity of countries to raise revenues, We suspect that the Spanish tax administration4 which includes an administration for the col- might have employed tax enforcement to react lection of taxes and the monitoring of compli- to regional sources of heterogeneity in the eco- ance (Besley and Persson 2009). From a dynamic nomic cycle. perspective, tax administration should then play This author also stresses negative features: a counter-cyclical role, that is, under a nega- it lacks transparency, it might be unfair and tive external shock (e.g., an economic downturn), distortionary, and—in a similar vein as a “tax tax enforcement should be reinforced.2 This idea amnesty”5 —it might make lose credibility to the seems to be confirmed in a related empirical study administration. That is why, Brondolo proposes a by Chen (2017) for China: a revenue loss (in that detailed practical strategy for tax administrations case, the abolition of a local tax) was largely off- to contain a potential rise in noncompliance dur- set by tougher tax enforcement. This evidence ing an economic crisis.6 confirms enforcement may be used by tax author- All in all, in front of an economic downturn ities as an instrument to counteract revenue losses the tax administration can react as follows: due to a negative shock. However, Chen considers an isolated shock, and the scarce empirical lit- (i) Strengthening tax enforcement, or tax erature does not address the potential existence capacity argument (Besley and Persson 2009). of a fiscal capacity argument in the setting of This would imply tax enforcement is counter- enforcement along the economic cycle. The con- cyclical and would be raised in order to over- text caused by the economic downturn associated come the financial constraints of public finances with the global financial crisis, with individu- in times of crisis. als and companies facing very important finan- (ii) Weakening tax enforcement. If the tax cial constraints, offers the opportunity to analyze administration prioritizes taxpayers’ welfare over the role of tax administration over the economic the revenue, enforcement would be procyclical as cycle.3 to smooth taxpayers’ financial shocks; or Brondolo (2009) argues tax enforcement has (iii) Tax enforcement is independent of the attractive features at stabilizing the economy, state of the economy. This is the implicit assump- tion made by the economic literature. 2. Kuehn (2014) provides a model for the informal econ- In this paper, we do not pretend to normatively omy in high-income countries and runs an interesting policy experiment with Spain and other Southern European coun- assess which one of the three options is the opti- tries: raising the enforcement standards to the Finnish level mal one, but to infer whether tax enforcement is (the best governance quality observed) would reduce the informal economy in Spain close to 8 percentage points and increase tax revenues by 7%. In order to provide a notion for 4. In Spain, tax administration tasks are shared by the the benefits of the policy, Kuehn assesses that the increase national tax agency (AEAT) and the tax agency of each in tax collection costs that would leave Spain indifferent Spanish region (the so-called autonomous communities—AC between adopting or not adopting the policy in terms of tax henceforth). The AEAT is responsible for the effective appli- revenue is 779%. Thus, the potential benefits of increasing cation of the main national taxes, such as personal income tax, enforcement in Spain would be large. value added tax and corporate income tax. Regional tax agen- 3. Almost 40% of the 49 revenue bodies analyzed by cies are responsible for the enforcement of annual wealth tax, the OECD, reported an increase in the aggregate value of inheritance and gift tax and tax on wealth transfers, in addition their debt inventory over the years 2007 to 2009 exceeding to other minor regional taxes. Therefore, the national agency 20%, and for 13 revenue bodies, this increase exceeded 40% plays a much more significant role than regional agencies. (OECD 2011). These are unpaid debts, that is, tax liabilities 5. See, though, Andreoni (1991) where he develops a recognized by taxpayers but not paid. After 2009 peak, aver- theoretical framework where a “tax amnesty” may increase age tax debt levels decreased, but in 2011 it remained in excess both efficiency and equity of the tax system. of 20% of the average reported for 2007. The incidence varied 6. Pappa, Sajedi, and Vella (2015) analyze how fiscal con- enormously across countries and in eight OECD countries the solidation is affected by the extent of evasion and corruption. level remained in 2011 over 50% their level in 2007 (OECD That is, they relate fiscal policy to compliance, but take this 2013). latter element as given. 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1877 influenced by economic cycles. Before moving to this question has remained unchanged over the the next section, where we will carefully explain 1994–2015 period. For any respondent i, in AC our identification procedure, note that the cycli- j, in survey year t, we code the answer to this cal characterization of enforcement might also question into the variable pijt ,which is scaled depend on how severe the economic crisis is. That from very low (1) to very high (4) according to is, the relationship between economic shock and the answer. Thus, by defining pijt as an ordinal enforcement could be nonlinear. We will also take dependent variable measuring the unobservable this into account in our empirical framework. actual perceived tax enforcement of individuals, p* ijt , we can design an ordered response model (see, e.g., Wooldridge 2002, 504–9)9 : III. EMPIRICAL ANALYSIS p∗ ijt = βECjt + Y ijt 𝛙 + Xjt 𝛂 + ϑj + τt + ϵijt The conceptual framework just described presents interesting and novel hypotheses about the evolution of tax enforcement along the ⎧ economic cycle that require empirical testing. ⎪1 if p∗ ijt ≤ ω1 Next, we present the employed methodology to ⎪2 if ω1 < p∗ ijt ≤ ω2 (1) pijt = ⎨ test this framework, discuss our identification ⎪3 if ω2 < p∗ ijt ≤ ω3 strategy, and finally present and comment on the ⎪4 if p∗ ijt > ω3 main results. ⎩ where ECjt is a proxy of the AC-specific eco- A. The Empirical Framework nomic cycle at time t. In particular, we alterna- tively employ the GDP of the AC j during year In order to test the hypotheses raised about t (GDPjt ) or the level of unemployment in AC the level of tax enforcement along the economic j during year t (Unemploymentjt ). Thus, a low cycle, we employ enforcement as it is perceived value of GDPjt with respect to its potential long- by individuals in Spain.7 This is our endogenous run level—implicitly accounted for through AC variable, which is extracted from the repeated fixed effects—implies an economic downturn. waves of the survey “Public opinion and fiscal The symmetric reasoning holds when Unemploy- policy,” conducted annually (1994–2015) and mentjt is employed as a proxy for the economic released by the Spanish Centre of Sociological cycle. Therefore, we identify a counter-cyclical Research (Centro de Investigaciones Sociológi- perceived tax enforcement with a negative (posi- cas in Spanish, CIS henceforth). This repeated tive) sign when ECjt is proxied by GDPjt (Unem- cross-section survey reports information on sub- ploymentjt ). jective perceptions of the fiscal policy, public pro- Since socioeconomic information about the vided goods and services, and other aspects of the respondents is also included in the survey, we tax system in Spain. collect this information in vector Y ijt to con- The relevant question used to define the trol for personal characteristics, while Xjt is a endogenous variable is the following: “Do vector collecting other AC-specific relevant vari- you think the tax administration is currently ables. We discuss in detail all these variables taking many/quite a few/a few/very few steps in its efforts to fight against tax evasion?”8 ; pocos o muy pocos esfuerzos para luchar contra el fraude fiscal?” (see e.g., question n. 21 of the survey n. 2,994 released 7. What matters about the decision to evade is the per- in 2013, as the numbering of the questions might change from ception of taxpayers (Slemrod 2019). For instance, Blank year to year). and Levin (2010) show that the U.S. Department of Jus- 9. Since the dependent variable is defined as an ordinal tice Tax Division issues a disproportionately large number of discrete ranking, employing an ordered response model is tax enforcement press releases during the weeks immediately the most appropriate estimation strategy. Indeed, as Greene prior to Tax Day (when income tax returns are due) compared (2002) states “although the outcome is discrete, the multino- to the rest of the year, with the aim to influence individual mial logit or probit model would fail to account for the ordinal taxpayers’ perceptions and knowledge of the audit probabil- nature of the dependent variable. Ordinary regression anal- ity. Hence, this perception might be directly affected by the ysis would err in the opposite direction, however. Take the actions of the tax administration. In any case, there is vast evi- outcome of an opinion survey. If the responses are coded 0, dence that individuals tend to overestimate the probability of 1, 2, 3, or 4, then linear regression would treat the difference their being audited even when fully informed about actual pol- between a 4 and a 3 the same as that between a 3 and a 2, icy (see e.g., Kahneman and Tversky 1979). We will explain whereas in fact they are only a ranking.” (see e.g., Greene in the next section how we tackle this issue. 2002, 736). However, as a robustness check we also repli- 8. The original question in Spanish is “¿Cree Ud. que, cate the analysis by estimating an OLS model (see the online en la actualidad. la Administración hace muchos, bastantes, Appendix S1). 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1878 ECONOMIC INQUIRY in Section III.B. Finally, we account for fixed effects (𝜗j ), time effects (𝜏 t ) and 𝜖 ijt is the error ⎧ ⎪1 if p∗ ijt ≤ ω1 term. We estimate the coefficients as well as ⎪2 if ω1 < p∗ ijt ≤ ω2 the cut-points in Equation (1) through an ordered (3) pijt = ⎨ . probit model10 by means of maximum likeli- ⎪3 if ω2 < p∗ ijt ≤ ω3 hood technique. ⎪4 if p∗ ijt > ω3 By estimating (1), we can evaluate the pooled ⎩ effect of the economic cycle on tax enforcement. Since nonlinear and linear terms are highly This effect could vary depending on the sever- correlated and there is the risk of getting ity of the economic shocks. In order to appreci- inflated standard errors, we orthogonalize the ate this potential change in the response of tax ECjt polynomial variables (Sribney 1995). enforcement, we employ a linear spline approach In the next section, we discuss our identifica- (see e.g., Poirier and Garber 1974; Gould 1993; tion strategy. or Johnston and DiNardo 1997) by specifying the relationship between p* ijt and ECjt as a piece- B. Identification Strategy wise seamless compound linear function. In other words, the relationship between p* ijt and ECjt Figure 1 shows the answers given by citizens is estimated as a function composed of linear to the question employed to define our endoge- segments that meet at the knots. The following nous variable. One can observe how this variable expression describes such specification: falls since 2009, that is, at the start of the finan- cial crisis with a lag and, then, goes up again p∗ ijt = f (ECjt ) + Y ijt 𝛙 + Xjt 𝛂 + ϑj + τt + ϵijt moderately from the beginning of the economic recovery. Thus, perceived enforcement shows a ⎧ β EC + a if EC ≤ knot cyclical pattern, in particular as a consequence of ⎪ 1 jt 1 jt 1 the Great Recession. f (ECjt ) = ⎨ β2 ECjt +a2 if knot1 knot Since we aim at estimating the potential tax ⎩ 3 jt 3 jt 2 enforcement response to asymmetric shocks at the regional level, we plot in Figure 2 each AC- ⎧ specific economic cycle11 vis-à-vis the average ⎪1 if p ijt ≤ ω1 ∗ evolution of our endogenous variable; we then ⎪2 if ω1 < p∗ ijt ≤ ω2 show the correlation coefficient between the two (2) pijt = ⎨ . ⎪ 3 if ω2 < p ijt ≤ ω3 ∗ series. These trends are obtained without control- ⎪4 if p∗ ijt > ω3 ling for any potential confounder and thus sim- ⎩ ple correlations cannot uncover a causal effect. The knots are alternatively equally spaced Nevertheless, we can observe variation across over the range of ECjt or are placed at convenient regions. In particular, in half of them there is a percentiles of ECjt in order to identify severe eco- prevailing procyclical trend, while in the rest the nomic downturns by considering extreme values trend is counter-cyclical. Given this first graph- of ECjt . Specifically, we set knot1 and knot2 at ical evidence, we aim at identifying the causal the first and the fifth percentiles of GDPjt or at impact of the cycle on enforcement in a robust the 95th and 99th percentiles of Unemploymentjt . way as we explain next. Alternatively, we also employ another standard Our endogenous variable is not a direct approach to identify nonlinearity that consists of description of the real efforts carried out by the including quadratic and cubic terms of ECjt in tax administration.12 Part of its variation along the regression model. This methodology is rep- resented by: 11. The AC-specific trend in the economic cycle is mea- sured as the residual part of the variation of the regional GDP p∗ ijt = β1 ECjt + β2 (ECjt )2 + β3 (ECjt )3 measured in logs which is not explained by the variation in logs of the national GDP and an AC-specific constant. + Y ijt 𝛙 + Xjt 𝛂 + ϑj + τt + ϵijt 12. There is no data on the number of audits performed by tax authorities at regional level. We have data on the total 10. The difference between an ordered probit and an number of employees of the tax administration in both AC- ordered logit model regards the distribution of 𝜀ijt . As main specific tax agencies and regional offices of the national tax strategy, by employing an ordered probit model, we are authority, but we cannot disaggregate this data by function. assuming a normal distribution of the error term. We also Thus, this data does not seem to be a good proxy for tax replicate the analysis assuming a logistic distribution, that is, enforcement but rather a measure of the size of each tax estimating an ordered logit model (see the Appendix S1). agency in any AC and year. Therefore, we employ this data to 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1879 FIGURE 1 Average pijt Variability along Time by AC. Note: y-axis represents the average score of pijt time can certainly be explained by how the actual one. Hence, since we want to identify the impact policy implemented by the administration varies of the latter on the actual policy, our strategy year-to-year. However, given the survey nature entails first controlling for the structural and com- of the variable, it might also vary along time due mon cyclical components of both dimensions of to the variation of individual risk perception even p* ijt , and second refining 𝛽 in order to identify if efforts carried out by the tax administration do the AC-specific cyclical component of the actual not change (see fn. 7); the response could also policy. Indeed, the coefficient 𝛽 in Equation (1) be normative, that is, what level of enforcement is picking up the potential effect of the AC- the interviewed thinks should hold at that par- specific economic cycle on both dimensions of ticular time. These last two dimensions are the p* ijt . The main challenge for a correct identifica- individual component of p* ijt ; in the regression, tion is being able to isolate such an effect. Below, we will include individual variables to control we detail the procedure we employ to deal with for those factors. this issue. Nonetheless, both the actual policy dimension and the individual one might be subject to dif- Controlling for the Actual Policy: Structural ferent types of fluctuations along the economic Component. In Equation (1), by employing fixed cycle. Indeed, both factors can be broken down effects (𝜗j ) and AC-specific contextual variables into a structural component, a common national (Xjt ), we are already implicitly controlling for cyclical component and an AC-specific cyclical the structural component of the actual policy. Nevertheless, we strengthen this strategy by define a variable controlling for the tax enforcement capacity controlling for 5 years fixed effects (i.e., by inter- at the regional level (see Section III.C for more details). Given acting the AC-specific dummies with 5 years this limitation, we think that individual perceptions of tax enforcement policies, once filtered by potential confounders common trend time dummies) instead of pure (i.e., normative beliefs and any particular bias due to risk fixed effects. In this way, we should control for aversion) play the right role as endogenous variable. potential changes in the long-term level of ECjt 1880 FIGURE 2 Trends in AC-Specific Economic Cycle Vis-À-Vis Average pijt Variability along Time by AC. ECONOMIC INQUIRY Note: left y-axis represents the log variation of the AC-specific Economic Cycle. This is measured as the residual part of the variation of the regional GDP measured in logs which is not explained by the variation in logs of the national GDP and an AC-specific constant; right y-axis represents the average score of pijt 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1881 over time.13 In order to account for AC-specific head of household, married individual, retired, contextual variables, vector Xjt includes several self-employed, public employee, left-wing voter, controls. First, we include a set of variables to nationalist voter,16 as well as for the estimated identify the regional productive structure through low unemployment risk (UR, henceforth).17 We the percentage composition of the regional gross also control for the municipality size,18 age of value added (GVA).14 Namely, the percentage the respondent—which are both included also of GVA represented by the secondary sector in squared terms to account for nonlinearity in (without the construction subsector), the per- their effect—and the educational level attained centage of GVA represented by the construction by the respondent. subsector—which has particularly been impor- tant in Spain—and finally the percentage of GVA Controlling for the Common Cyclical Compo- given by the tertiary sector.15 We also include the nents (Individual Dimension and Actual Policy). regional population in order to account for the By employing common time effects (𝜏 t ), we demographic dimension of any AC, and the total account for the common national cyclical com- number of employees of the tax administration in ponent of our endogenous variable. Moreover, in per capita terms in order to account for the capac- order to control for the sensitivity of the individ- ity of the tax authority to enforce the existing ual component of p* ijt on the national economic tax legislation. The regional deficit is included cycle, we interact the vector of individual vari- to synthetically account for budgetary shocks on ables (Y ijt ) with the common time dummies (𝜏 t ) both government expenditure and revenues and allowing the risk perception and demand dimen- to further isolate these shocks from the business sion of p* ijt to sluggishly adjust along the national cycle. Finally, we include a dummy variable to economic cycle. Indeed, respondents may answer account for the AC electoral cycle and a dummy differently implicitly showing a different risk identifying whether the AC government stands perception/demand of tax enforcement efforts on the left of the political spectrum. depending on the national economic cycle. Controlling for the Individual Dimension: Controlling for the Individual Dimension: AC- Structural Component. In order to account Specific Cyclical Component. Despite account- for the structural component of the individual ing for the aforementioned effects, serial corre- dimension of perceived tax enforcement, we lation might still be present. Indeed, the relation- control for individual characteristics of the ship between the AC-specific economic cycle and respondents (Y ijt ), which may influence the our endogenous variable will capture both the risk perception and the enforcement demand of sensitivity of the tax administration throughout individuals. We include dummies for female, the economic cycle (i.e., the AC-specific cyclical component of the actual policy) and the evolution 13. Alternatively, we also employ as a robustness 3- and of individual risk perception/demand throughout 10-year fixed effects. The results of this robustness analysis the cycle (i.e., the AC-specific cyclical compo- are reported in the Appendix S1. nent of the individual dimension of p* ijt ). This 14. These are introduced with 5-year lags in order to account for the long-run productive structure of any AC. The means that the estimated effect of the cycle on regional composition of the GVA does not vary much along the (latent) endogenous variable—the coefficient the analyzed period and controlling for the GVA structure 𝛽 —could be biased. without lags does not qualitatively alter the results which are available upon request. In order to deal with this issue, we follow 15. More precisely, the primary sector includes agricul- the approach adopted by Backus and Esteller- ture, forestry, and fishing. The secondary sector is defined Moré (2017). The initial step of this strategy is to to include extractive industries; manufacturing industries; energy industries (electricity, gas, steam, and air condition- ing); water supply; sanitation, waste management, and decon- 16. The dummy nationalist is defined as equal to 1 if tamination. As mentioned in the text, the construction indus- the respondent voted for one of the nationalist parties of the try is considered as a subsector a part. Finally, the tertiary historical nationalities recognized in Spain. sector includes the following services: wholesale and retail 17. The methodology employed to estimate UR is trade; repair of motor vehicles and motorcycles; transport explained below in this section. and storage; hospitality; information and communications; 18. The municipality size is coded as to rank from financial and insurance activities; real estate activities; pro- 0 to 7 in this way: 0—less or equal to 1,000 inhabi- fessional, scientific, and technical activities; administrative tants; 1—1,001 to 2,000 inhabitants; 2—2,001 to 10,000 activities and auxiliary services; public administration and inhabitants; 3—10,001 to 50,000 inhabitants; 4—50,001 defense; compulsory social security; education, health, and to 100,000 inhabitants; 5—100,001 to 400,000 inhabi- social services activities; artistic, recreational, and entertain- tants; 6—400,001 to 1,000,000 inhabitants; 7—more than ment activities; repair of household items; and other services. 1,000,000 inhabitants. 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1882 ECONOMIC INQUIRY split our sample of surveyed individuals into two choose ̂ β1 as the best approximation to the real groups, the first one—say group 1—composed impact of the AC economic cycle on the actual tax by people whose risk perception and demand of enforcement policy. Otherwise, we can conclude tax enforcement should not vary along the AC- that this source of bias is not relevant. specific economic cycle and the second one con- As a final step of our identification strategy, we stituted by the complementary cluster (group 2). perform a falsification test aimed at determining To this end, coherently with Backus and Esteller- the reliability of the results based on the filtering Moré (2017), we provide an estimate of each indi- process presented above. To this end, we employ vidual’s UR based on their labor market char- the same identification strategy to test cyclical- acteristics.19 This is an estimate of an individ- ity in tax policy. Similarly, in order to define the ual’s idiosyncratic risk of unemployment, scaled dependent variable, we employ another question between 0 and 1, and provides us with a proxy of of the CIS’ repeated surveys: “Would you say that the impact of the economic cycle on the risk per- what the Spaniards pay in taxes is a lot, regu- ception/demand of tax enforcement of those indi- lar or little?”22 ; again, this question has remained viduals. The rationale is that the higher the UR, unchanged over the 1994–2015 period. Analo- the higher the individual’s exposure to the eco- gously to what we do for pijt , we code the answers nomic cycle, and thus, the higher should be the to this question into an ordered variable tijt ,which potential impact of the economic cycle on her risk is scaled from “low tax pressure” (1) to “high tax perception and probably demand for tax enforce- pressure” (3). This is a measure of the perceived ment. In other words, an individual with low UR tax pressure t* ijt . By employing the same filter- is less likely to change her perception/demand of ing process, we then estimate a model analogous tax enforcement along the economic cycle and to Equation (1), but employing tijt as dependent thus is more likely to contribute to produce a cor- variable. If our identification strategy is prop- rect estimation of 𝛽. We identify the cluster of erly filtering the impact of the AC-specific eco- individuals whose risk perception and demand of nomic cycle on the actual policy, we would expect tax enforcement should not vary along the AC results coherent with the hypothesis of acyclical economic cycle (group 1) by defining the dummy tax policy. Indeed, while the individual compo- variable “low UR” equal to 1 if the UR of a cer- nent of t* ijt may well be either pro or counter- tain individual in year t is lower than the average cyclical according to changes in the individual UR of that year.20 risk perception or demand, it is unlikely policy- The following step consists of running sepa- makers react to the AC-specific economic cycle rate regressions for these two groups, and check- when it comes to statutory tax policy. This means ing whether there is a statistically significant dif- that, after having filtered t* ijt for all the individ- ference between the estimated coefficient ̂ β for ual confounders as well as for the structural and the two clusters (i.e., whether ̂ β1 ≠ ̂β2 is statisti- common cyclical components, we expect a non- cally significant21 ). If this is the case, and accord- statistically significant impact of the AC-specific ing to our identification strategy, then we should economic cycle on tax policy. This would corrob- orate our identification strategy. 19. More precisely, we estimate the individual UR by employing a probit model on a subsample of individu- In the next section, we present some descrip- als who are employed plus those that are currently unem- tive statistics and detail the sources of the variable ployed but were employed in previous periods. Specifically, included in the analysis. we establish the relationship URijt = wijt 𝝎 + 𝜂 ijt where URijt is a dummy equal to 1 if i is unemployed and 0 if i is employed, wijt is a vector of i’s employment—or previous C. Data and Sources employment—characteristics reported in the CIS surveys. Those include occupation, industry of employment, and level Our data set comprises information about of education all interacted with the sector of employment and individual-level and AC-level variables for the year effects, 𝝎 is a vector of parameters to be estimated, and 1994–2015 period. Our dependent variable, ̂ ijt , repre- 𝜂 ijt is the error term. The predicted probabilities UR including the one employed for the falsification sent the estimated UR variable (for more details see Backus and Esteller-Moré 2017, 207). Additionally, we assign a value test, and all the individual-level control variables, equal to zero to the UR of retired individuals. 20. Alternatively, we have also employed the median of to group 1 or 2 (see Backus and Esteller-Moré 2017, 209 for the UR of any year to define the threshold and obtain qualita- more details). tively the same results, which are available upon request. 22. The original question in Spanish is “¿Diría Ud. que 21. In order to perform this test, we implement a fully lo que los/as españoles/as pagamos en impuestos es mucho, interacted specification (FIS) of Equation (2), by allowing any regular o poco?” (see e.g., question n. 11 of the survey n. 2994 coefficient to differ depending on whether individual i belongs released in 2013). 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1883 TABLE 1 Pooled Summary Statist ics Variable Measurement unit Obs. Mean SD Min Max Dependent variable p Ranking 32,357 2.360 0.812 1 4 Falsification test dependent variable t Ranking 35,152 2.607 0.542 1 3 Proxies of the Economic Cycle (main explanatory variables) GDP (CA) Hundreds of billions of 36,935 1.006 0.643 0.051 2.150 euros Unemployment (CA) Millions of people 36,935 0.427 0.425 0.005 2.186 AC-specific explanatory variables Deficit (CA) Hundreds of billions of 34,345 −0.010 0.021 −0.107 0.015 euros FYL_%GVA_Secondary_sector Share 36,935 0.219 0.073 0.076 0.441 FYL_%GVA_Construction_sector Share 36,935 0.092 0.029 0.029 0.149 FYL_% GVA_Tertiary_sector Share 36,935 0.647 0.076 0.416 0.827 Population (CA) Millions of people 34,345 4.606 2.488 0.263 8.450 Leftist government (CA) Dummy 36,935 0.352 0.478 0 1 Electoral cycle (CA) Dummy 36,935 0.277 0.447 0 1 TA employees per thousands of people Employees per thousands 32,522 0.725 0.225 0.343 1.615 of people Individual-level explanatory variables Dummy self-employed Dummy 36,946 0.147 0.354 0 1 Left Dummy 36,946 0.535 0.499 0 1 Female Dummy 36,946 0.422 0.494 0 1 Age Nr. of years 36,940 48.899 17.994 18 99 Age squared Nr. of years (squared) 36,9402, 714.857 1872.325 324 9, 801 Head of household Dummy 36,946 0.609 0.488 0 1 Dummy married Dummy 36,919 0.353 0.478 0 1 Education level Nr. of years 36,852 4.296 3.042 0 15 Nationalist Dummy 36,946 0.065 0.246 0 1 Municipality size Units 36,946 3.351 2.126 0 7 Municipality size squared Units squared 36,946 15.748 15.216 0 49 Dummy retired Dummy 36,946 0.329 0.470 0 1 Dummy public employee Dummy 36,946 0.171 0.376 0 1 Unemployment risk Probability 35,369 0.109 0.144 0 0.837 Dummy low Unemployment risk (mean) Dummy 36,946 0.621 0.485 0 1 Note: The dependent variables and all the individual-level control variables, are extracted from the repeated surveys waves of the “Public opinion and public policy” annually published by the CIS. UR, and the relative dummy for low unemployment risk have been estimated (see Section III.B). The GDP, the unemployment, the productive structure and the population of ACs is provided by the INE. To define the total number of employees of the tax administration we rely both on information provided by the Statistical Bulletins of the Central Personnel Registry and on information made available in the Report on the ceded taxes to ACs published every year jointly with the project of the general State budget. Information on the electoral cycle and on the political color of the government in office in any AC/year is available on the database of the Spanish Interior Ministry. are extracted from the abovementioned repeated for the number of employees employed in both waves of the annually published survey by the AC-specific tax agencies and regional offices of CIS. The only exception is UR, and the related the national tax authority and it is relativized per dummy for low unemployment risk, which were capita terms. This variable represents a measure estimated using the methodology presented in of the size of each tax agency with respect to the Section III.B. population in any AC and year. In order to define Contextual variables refer to the 15 Spanish this variable, we rely both on information pro- “common regime” ACs and are obtained from vided by the Statistical Bulletins of the Central the following statistical sources. The informa- Personnel Registry (“Boletines Estadísticos del tion about the GDP, the unemployment, the pro- Registro Central de Personal” in Spanish) and on ductive structure, and the population of ACs is information made available in the Report on the provided by the Spanish National Institute of ceded taxes to ACs (“Informe sobre la cesión de Statistics (INE). The variable that controls for the tributos a las Comunidades Autónomas” in Span- tax enforcement capacity—the total number of ish) published every year jointly with the project employees of the tax administration—accounts of the general State budget. Information on the 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1884 ECONOMIC INQUIRY TABLE 2 Summary Statistics by UR Type Panel A: High UR Variable Measurement Unit Obs. Mean SD Min Max Dependent variable p Ranking 12,593 2.308 0.813 1 4 Falsification test dependent variable t Ranking 13,478 2.632 0.531 1 3 Proxies of the Economic Cycle (main explanatory variables) GDP (CA) Hundreds of billions of euros 13,984 1.020 0.638 0.051 2.150 Unemployment (CA) Millions of people 13,984 0.432 0.417 0.005 2.186 AC-specific explanatory variables Deficit (CA) Hundreds of billions of euros 12,987 −0.011 0.022 −0.107 0.015 FYL_%GVA_Secondary_sector Share 13,984 0.219 0.074 0.076 0.441 FYL_%GVA_Construction_sector Share 13,984 0.088 0.032 0.029 0.149 FYL_% GVA_Tertiary_sector Share 13,984 0.652 0.075 0.416 0.827 Population (CA) Millions of people 12,987 4.691 2.492 0.263 8.450 Leftist government (CA) Dummy 13,984 0.341 0.474 0 1 Electoral cycle (CA) Dummy 13,984 0.321 0.467 0 1 TA employees per thousands of people Employees per thousands of 12,281 0.711 0.218 0.343 1.515 people Individual-level explanatory variables Dummy self-employed Dummy 13,990 0.111 0.314 0 1 Left Dummy 13,990 0.570 0.495 0 1 Female Dummy 13,990 0.425 0.494 0 1 Age Nr. of years 13,989 39.082 13.457 18 99 Age squared Nr. of years (squared) 13,989 1, 708.481 1, 188.069 324 9, 801 Head of household Dummy 13,990 0.516 0.500 0 1 Dummy married Dummy 13,980 0.419 0.493 0 1 Education level Nr. of years 13,971 4.495 4.951 0 99 Nationalist Dummy 13,990 0.060 0.237 0 1 Municipality size Units 13,990 3.381 2.018 0 7 Municipality size squared Units squared 13,990 15.505 14.465 0 49 Dummy Retired Dummy 13,990 0.051 0.219 0 1 Dummy public employee Dummy 13,990 0.108 0.310 0 1 Unemployment risk Probability 12,413 0.254 0.155 0 0.837 Panel B: Low UR Variable Measurement Unit Obs. Mean SD Min Max Dependent variable p Ranking 19,764 2.393 0.809 1 4 Falsification test dependent variable t Ranking 21,674 2.591 0.548 1 3 Proxies of the Economic Cycle (main explanatory variables) GDP (CA) Hundreds of billions of euros 22,951 0.997 0.646 0.051 2.150 Unemployment (CA) Millions of people 22,951 0.423 0.431 0.005 2.186 AC-specific explanatory variables Deficit (CA) Hundreds of billions of euros 21,358 −0.010 0.020 −0.107 0.015 FYL_%GVA_Secondary_sector Share 22,951 0.220 0.073 0.076 0.441 FYL_%GVA_Construction_sector Share 22,951 0.094 0.026 0.030 0.149 FYL_% GVA_Tertiary_sector Share 22,951 0.644 0.076 0.416 0.827 Population (CA) Millions of people 21,358 4.555 2.485 0.263 8.450 Leftist government (CA) Dummy 22,951 0.360 0.480 0 1 Electoral cycle (CA) Dummy 22,951 0.250 0.433 0 1 TA employees per thousands of people Employees per thousands of 20,241 0.733 0.229 0.343 1.615 people Individual-level explanatory variables Dummy self-employed Dummy 22,956 0.169 0.375 0 1 Left Dummy 22,956 0.514 0.500 0 1 Female Dummy 22,956 0.421 0.494 0 1 Age Nr. of years 22,951 54.882 17.784 18 99 Age squared Nr. of years (squared) 22,951 3, 328.259 1946.403 324 9, 801 Head of household Dummy 22,956 0.665 0.472 0 1 Dummy married Dummy 22,939 0.312 0.463 0 1 Education level Nr. of years 22,951 4.464 5.220 0 99 Nationalist Dummy 22,956 0.068 0.252 0 1 Municipality size Units 22,956 3.332 2.189 0 7 Municipality size squared Units squared 22,956 15.896 15.654 0 49 Dummy Retired Dummy 22,956 0.498 0.500 0 1 Dummy public employee Dummy 22,956 0.209 0.406 0 1 Unemployment risk Probability 22,956 0.031 0.039 0 0.170 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1885 TABLE 2 Continued Panel C: Means Difference High UR – Low UR Variable Mean High UR Mean Low UR Difference p Value Dependent variable p 2.308 2.393 −0.085 .000*** Falsification test dependent variable t 2.632 2.591 0.041 .000*** Proxies of the Economic Cycle (main explanatory variables) GDP (CA) 1.020 0.997 0.023 .001*** Unemployment (CA) 0.432 0.423 0.009 .040** AC-specific explanatory variables Deficit (CA) −0.011 −0.010 −0.001 .000*** FYL_%GVA_Secondary_sector 0.219 0.220 −0.001 .312 FYL_%GVA_Construction_sector 0.088 0.094 −0.006 .000*** FYL_% GVA_Tertiary_sector 0.652 0.644 0.008 .000*** Population (CA) 4.691 4.555 0.136 .000*** Leftist government (CA) 0.341 0.360 −0.019 .000*** Electoral cycle (CA) 0.321 0.250 0.071 .000*** TA employees per thousands of people 0.711 0.733 −0.022 .000*** Individual-level explanatory variables Dummy self-employed 0.111 0.169 −0.058 .000*** Left 0.570 0.514 0.056 .000*** Female 0.425 0.421 0.005 .365 Age 39.082 54.882 −15.800 .000*** Age squared 1,708.481 3,328.259 −1,619.778 .000*** Head of household 0.516 0.665 −0.150 .000*** Dummy married 2.0000 0.419 0.312 .107 Education level 4.495 4.464 0.031 .576 Nationalist 0.060 0.068 −0.008 .002*** Municipality size 3.381 3.332 0.049 .032** Municipality size squared 15.505 15.896 −0.391 .017** Dummy retired 0.051 0.498 −0.447 .000*** Dummy public employee 0.108 0.209 −0.101 .000*** Unemployment risk 0.254 0.031 0.223 .000*** Note: The dependent variables and all the individual-level control variables, are extracted from the repeated surveys waves of the “Public opinion and public policy” annually published by the CIS. UR, and the relative dummy for low unemployment risk have been estimated (see Section III.B). The GDP, the unemployment, the productive structure and the population of ACs is provided by the INE. To define the total number of employees of the tax administration we rely both on information provided by the Statistical Bulletins of the Central Personnel Registry and on information made available in the Report on the ceded taxes to ACs published every year jointly with the project of the general State budget. Information on the electoral cycle and on the political color of the government in office in any AC/year is available on the database of the Spanish Interior Ministry. *p < .10; **p < .05; ***p < .01. electoral cycle and on the political color of the The structure of the table is coherent with the government in office in any AC/year is available filtering process presented in our estimation strat- on the database of the Spanish Interior Ministry. egy. More precisely, columns 1 and 4 estimate the Table 1 reports the pooled summary statistics of baseline model presented in Equation (1) includ- the variables employed in our empirical analy- ing fixed and time effects, columns 2 and 5 sub- sis, while Table 2 presents the descriptive statis- stitute standard fixed effects with 5-year fixed tics referred to the two subsamples defined on the effects and finally in columns 3 and 6, we add basis of the dummy “low UR” (panels A and B) the interaction between any individual variable and a test for the equality of subsamples means and the time dummies. In every model, the proxy (Panel C). for the economic cycle is highly significant and Next, we present the results of our empiri- presents a sign that is coherent with a pooled cal analysis. counter-cyclical perceived tax enforcement. Regarding our filtering process, by substitut- IV. RESULTS ing standard fixed effects with 5-year fixed effects has a significant impact on the magnitude of the A. Main Results coefficients of the economic cycle, while intro- Table 3 presents the results of the estimation ducing the interactions between individual vari- of Equation (1). In particular, in columns 1–3, ables and time dummies has a negligible if not we measure the economic cycle through GDPjt , null impact. This seems to suggest that the indi- while in columns 4–6, we use Unemploymentjt . vidual component of the perception/demand of 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1886 ECONOMIC INQUIRY TABLE 3 The Determinants of Perceived Tax Enforcement Along Time. Ordered-Probit, 1994–2015 (1) (2) (3) (4) (5) (6) GDP (CA) −0.409*** −0.788*** −0.770*** (−3.695) (−3.861) (−3.622) Unemployment (CA) 0.180*** 0.202*** 0.202*** (4.361) (3.365) (3.126) Deficit (CA) 1.901** 1.006 1.138 1.663** 1.270 1.387 (2.396) (0.814) (0.885) (2.171) (1.029) (1.080) FYL_%GVA_Secondary_sector 1.327 −0.590 −0.707 0.200 −1.955 −1.923 (1.469) (−0.417) (−0.470) (0.226) (−1.327) (−1.238) FYL_%GVA_Construction_sector −1.292 0.983 1.372 −1.049 0.590 1.126 (−1.163) (0.494) (0.661) (−0.952) (0.296) (0.540) FYL_% GVA_Tertiary_sector 2.450** 2.718* 2.668* 1.860* 2.010 2.077 (2.452) (1.750) (1.653) (1.875) (1.271) (1.269) Leftist government (CA) 0.011 0.024 −0.008 0.017 0.047 0.013 (0.448) (0.638) (−0.203) (0.717) (1.233) (0.345) Electoral cycle (CA) −0.004 −0.003 −0.005 −0.003 0.003 0.001 (−0.209) (−0.177) (−0.276) (−0.159) (0.148) (0.076) TA employees per thousands of people −0.412*** −0.398*** −0.439*** −0.399*** −0.390*** −0.432*** (−7.048) (−6.603) (−7.192) (−6.813) (−6.450) (−7.060) Population (CA) 0.165*** 0.223 0.190 −0.011 −0.061 −0.077 (2.659) (1.567) (1.272) (−0.281) (−0.486) (−0.578) Municipality size −0.086*** −0.085*** −0.033 −0.087*** −0.085*** −0.033 (−4.719) (−4.607) (−0.369) (−4.752) (−4.623) (−0.368) Municipality size squared 0.009*** 0.009*** 0.005 0.009*** 0.009*** 0.005 (4.048) (3.854) (0.423) (4.075) (3.860) (0.422) Left −0.037*** −0.035*** −0.195*** −0.037*** −0.034*** −0.195*** (−2.772) (−2.596) (−2.854) (−2.758) (−2.583) (−2.854) Female −0.008 −0.007 −0.065 −0.008 −0.007 −0.065 (−0.562) (−0.496) (−0.946) (−0.553) (−0.497) (−0.946) Age −0.005** −0.004* −0.018 −0.005** −0.004* −0.018 (−1.965) (−1.929) (−1.451) (−1.977) (−1.920) (−1.451) Age squared 0.000*** 0.000*** 0.000 0.000*** 0.000*** 0.000 (2.802) (2.766) (1.505) (2.815) (2.766) (1.506) Head of household 0.013 0.015 −0.009 0.013 0.015 −0.009 (0.813) (0.982) (−0.123) (0.844) (0.984) (−0.122) Dummy married −0.023 −0.024 −0.033 −0.023 −0.023 −0.032 (−1.518) (−1.555) (−0.366) (−1.497) (−1.518) (−0.364) Dummy self-employed 0.044* 0.045** 0.109 0.043* 0.044* 0.109 (1.925) (1.961) (0.961) (1.894) (1.937) (0.962) Dummy Retired −0.015 −0.013 −0.223 −0.015 −0.014 −0.223 (−0.576) (−0.520) (−1.617) (−0.574) (−0.536) (−1.618) Dummy public employee −0.032* −0.029 0.018 −0.032* −0.029 0.018 (−1.728) (−1.564) (0.193) (−1.714) (−1.564) (0.194) Dummy low Unemployment risk (mean) 0.044** 0.042** 0.136 0.044** 0.042** 0.136 (2.341) (2.240) (1.443) (2.334) (2.243) (1.442) Education level 0.008*** 0.008*** 0.001 0.008*** 0.008*** 0.001 (5.161) (5.196) (0.165) (5.137) (5.181) (0.164) Nationalist 0.076*** 0.069** −0.009 0.077*** 0.071** −0.009 (2.650) (2.404) (−0.059) (2.664) (2.449) (−0.059) Observations 28,384 28,384 28,384 28,384 28,384 28,384 Log-likelihood −32,875.604 −32,793.128 −32,554.447 −32,872.955 −32,795.524 −32,556.497 Fixed effects Yes No No Yes No No Time effects Yes Yes Yes Yes Yes Yes FE ×5 years TE No Yes Yes No Yes Yes Individual_Var.s × TE No No Yes No No Yes Note: z statistics are reported in parentheses and are derived using robust standard errors. We do not employ standard errors clustered at regional level, since we have too many regressors in comparison to clusters. The theory that justifies the calculation of cluster standard errors is asymptotic in the number of clusters, and the established rule is to employ a number of clusters that at least equates the number of estimated variables. Otherwise, one needs to consider whether any of the reported standard errors means anything. However, we use Fixed Effects (or even 5 years FE) at regional level which control for most of the within-cluster correlation of the error. *p < .10; **p < .05; ***p < .01. 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1887 tax enforcement does not vary much along the of enforcement (generally corroborating the economic cycle. For these reasons, and in order results obtained by Chen 2017), but as a novelty to be able to more easily interpret the effect of rel- it relates this result to economic cyclical shocks. evant individual variables, we choose the results Indeed, our results implicitly suggest that, on for 5-year fixed effects (columns 2 and 5) as our average, tax revenue losses due to the economic best estimates for the pooled regression. Further downturn tend to be accompanied by perceived analyses presented in Tables 4 and 5 are based on tougher enforcement. Next, we try to disentangle these models. whether the severity of the economic downturn As explained in detail in Section III.B, the affects perceived tax enforcement. control variables have been included as part of In this vein, Table 5 presents the results of the our identification strategy. Thus, the interpreta- analysis of the presence of potential nonlinear- tion of their impact on the dependent variable ity. More specifically, columns 1–4 are related to is not key for the purpose of this paper. Nev- Equation (2). Columns 1 and 2 employ a linear ertheless, it is interesting to stress some results. spline methodology with equally spaced knots; In particular, regions with a higher percentage of columns 3 and 4 use a linear spline methodology GVA generated by the tertiary sector tend to have with knots at specified extreme points (i.e., first higher perceived tax enforcement. The capac- and fifth percentiles for the GDP based model ity of tax administration to enforce the exist- and 95th and 99th percentiles for the unemploy- ing tax legislation seems to be oversized as the ment based model); and columns 5 and 6 present coefficient of the TA employees per thousands the results of the estimation of Equation (3) that of people variable suggests. The impact of size employs an orthogonalized third degree poly- of the municipality in which the respondent to nomial to account for nonlinearity in the eco- the survey resides is reported to be nonlinear as nomic cycle. the impact of the age of the respondent. Left- These results seem to suggest a change in ist voters report a lower perceived level of tax the perceived behavior of tax administration. enforcement suggesting a demand for a more Namely, the models that employ linear spline stringent fight against fiscal fraud. On the other with equally spaced knots do not show a change hand, being a voter of a regional nationalist party in the sign of the slope, but we can at least appre- has the opposite effect on the individual percep- ciate a change in the slope magnitude. A draw- tion/demand. Self-employed individuals report a back of this approach is that in order to iden- higher level of perceived tax enforcement which tify the change in the economic cycle employ- is coherent with the higher probability they have ing knots, they are equally spaced. By using lin- to be audited. Finally, individuals employed in ear spline models with knots at specified extreme the public sector tend to report a lower level points that identify severe financial constraints, of perceived tax enforcement, while people with though, we are able to appreciate a significant a lower estimated expected unemployment rate change in the slope of perceived tax enforcement show the opposite effect. to the economic cycle. More specifically, column Following our identification strategy, Table 4 3 (4) show that for very low (high) values of presents the results of the estimation of sepa- GDP (unemployment) taxpayers’ perception of rate regressions for different groups of individu- tax enforcement policy turns out to be procycli- als based on their UR-type. In particular, columns cal while remaining counter-cyclical for the rest 1 and 2 replicate model 2 of Table 3 for low-UR of the economic cycle. We obtain a similar effect type and high-UR type individuals, respectively. also for the results related to Equation (3) but Similarly, columns 3 and 4 reproduce model 5 just for what concerns the unemployment-based of Table 3 for the same clusters of individu- model (column 6). als. The results show perceived counter-cyclical In the online appendix, we present the tax enforcement for both clusters of individuals. results of our robustness analyses. In particu- Testing for significantly different coefficients for lar, Tables A1 to A6, Supporting Information, these two groups lead to rejecting this hypothe- replicate Tables 3–5 by alternatively employing sis, so we maintain the results shown in model 3 years or 10 years fixed effects instead of 5 years 2 and 5 of Table 3 as best approximation to the fixed effects in order to account for potentially pooled effect of the economic cycle on perceived different economic cycles. Results remain qual- tax enforcement. itatively unchanged. Online Tables A7 to A12 Hence, thus far, our study confirms the exis- present the results obtained by replicating the tence of a fiscal capacity argument in the setting analysis presented in Tables 3–5 by estimating 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1888 ECONOMIC INQUIRY TABLE 4 The Determinants of Perceived Tax Enforcement Along Time. Ordered-Probit, 1994–2015; Separate Regressions by UR Type (1) (2) (3) (4) Low UR High UR Low UR High UR GDP (CA) −0.782*** −0.793** (−3.038) (−2.282) Unemployment (CA) 0.171** 0.241** (2.286) (2.311) Deficit (CA) −0.752 3.409* −0.501 3.724* (−0.463) (1.755) (−0.309) (1.925) FYL_%GVA_Secondary_sector −1.131 1.618 −2.158 −0.384 (−0.665) (0.607) (−1.222) (−0.138) FYL_%GVA_Construction_sector −1.339 5.774 −1.691 5.338 (−0.546) (1.625) (−0.687) (1.498) FYL_% GVA_Tertiary_sector 3.366* 2.907 2.779 1.989 (1.775) (1.019) (1.437) (0.685) Leftist government (CA) −0.010 0.099 0.010 0.121* (−0.220) (1.568) (0.200) (1.930) Electoral cycle (CA) 0.005 −0.026 0.011 −0.021 (0.195) (−0.799) (0.459) (−0.652) TA employees per thousands of people −0.386*** −0.433*** −0.378*** −0.423*** (−5.042) (−4.328) (−4.924) (−4.223) Population (CA) 0.029 0.467** −0.232 0.166 (0.149) (1.993) (−1.337) (0.790) Municipality size −0.087*** −0.069** −0.087*** −0.068** (−3.734) (−2.258) (−3.757) (−2.239) Municipality size squared 0.009*** 0.007* 0.009*** 0.007* (3.149) (1.850) (3.154) (1.834) Left −0.050*** −0.012 −0.050*** −0.012 (−2.935) (−0.561) (−2.927) (−0.540) Female −0.015 0.003 −0.015 0.003 (−0.799) (0.120) (−0.791) (0.113) Age −0.007** −0.001 −0.007** −0.001 (−2.125) (−0.254) (−2.116) (−0.234) Age squared 0.000*** 0.000 0.000*** 0.000 (3.079) (0.391) (3.073) (0.376) Head of household 0.015 0.013 0.015 0.012 (0.771) (0.507) (0.770) (0.497) Dummy married −0.002 −0.058** −0.002 −0.057** (−0.113) (−2.514) (−0.105) (−2.463) Dummy self-employed 0.038 0.069 0.038 0.068 (1.449) (1.341) (1.426) (1.325) Dummy retired −0.032 0.054 −0.032 0.052 (−1.004) (0.773) (−1.009) (0.743) Dummy public employee −0.034 −0.005 −0.034 −0.004 (−1.551) (−0.141) (−1.564) (−0.112) Education level 0.007*** 0.008*** 0.007*** 0.008*** (3.583) (3.703) (3.570) (3.696) Nationalist 0.091** 0.031 0.091** 0.033 (2.477) (0.641) (2.482) (0.689) Observations 17,371 11,013 17,371 11,013 Log-likelihood −20,002.242 −12,705.151 −20,004.539 −12,705.350 Fixed Effects No No No No Time Effects Yes Yes Yes Yes FE ×5 years TE Yes Yes Yes Yes Individual_Var.s × TE No No No No Note: z statistics are reported in parentheses and are derived using robust standard errors. *p < .10; **p < .05; ***p < .01. TABLE 5 The Determinants of Perceived Tax Enforcement Along Time. Ordered-Probit, 1994–2015; Nonlinearity in the Response to Economic Cycle (1) (2) (3) (4) (5) (6) Linear Spline Linear Spline Orthogonalized with Knots with Knots At Third Degree Equally Spaced Specified Points Polynomial (first and fifth pctls) (95th and 99th pctls) GDP (CA)1 −0.997*** 10.637* (−3.099) (1.821) GDP (CA)2 −0.337** −3.260 (−2.363) (−1.078) GDP (CA)2 −0.503*** −0.418*** (−4.295) (−4.267) Unemployment (CA)1 −0.160 0.184* (−0.968) (1.728) Unemployment (CA)2 0.369*** 0.522** (2.727) (2.306) Unemployment (CA)3 −0.011 −4.958*** (−0.067) (−2.740) Orthogonalized GDP (CA) −0.088 (−1.249) Orthogonalized [GDP (CA)]2 −0.463*** (−3.297) Orthogonalized [GDP (CA)]3 −0.013 (−0.487) Orthogonalized Unemployment (CA) 0.269** (2.487) Orthogonalized [Unemployment (CA)]2 0.231*** (2.697) Orthogonalized [Unemployment (CA)]3 −0.231** (−2.443) Observations 28,384 28,384 28,384 28,384 28,384 28,384 Log-likelihood −32,872.263 −32,791.689 −32,873.986 −32,791.381 −32,792.650 −32,791.272 AC-specific explanatory variables Yes Yes Yes Yes Yes Yes Individual-level explanatory variables Yes Yes Yes Yes Yes Yes Fixed effects No No No No No No Time effects Yes Yes Yes Yes Yes Yes DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT FE × 5 years TE Yes Yes Yes Yes Yes Yes Individual_Var.s × TE No No No No No No Note: z statistics are reported in parentheses and are derived using robust standard errors. *p < .10; **p < .05; ***p < .01. 1889 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1890 ECONOMIC INQUIRY TABLE 6 Falsification Test—The Determinants of Perceived Tax Pressure Along Time. Ordered-Probit, 1994–2015 (1) (2) (3) (4) (5) (6) GDP (CA) 0.256** −0.050 −0.097 (2.084) (−0.230) (−0.423) Unemployment (CA) −0.134*** −0.043 −0.078 (−2.813) (−0.634) (−1.070) Deficit (CA) 1.563* 1.983 1.446 1.600* 1.971 1.434 (1.799) (1.454) (1.015) (1.915) (1.445) (1.006) FYL_%GVA_Secondary_sector −0.727 −0.687 −1.808 0.005 −0.404 −1.325 (−0.738) (−0.446) (−1.100) (0.005) (−0.253) (−0.782) FYL_%GVA_Construction_sector 0.754 −0.456 −2.329 0.613 −0.321 −2.040 (0.633) (−0.211) (−1.024) (0.522) (−0.148) (−0.892) FYL_% GVA_Tertiary_sector −0.706 −1.195 −2.511 −0.342 −0.950 −2.105 (−0.642) (−0.689) (−1.383) (−0.314) (−0.538) (−1.142) Leftist government (CA) −0.032 −0.019 −0.003 −0.037 −0.023 −0.009 (−1.173) (−0.464) (−0.068) (−1.362) (−0.540) (−0.199) Electoral cycle (CA) 0.005 −0.003 0.006 0.005 −0.001 0.010 (0.229) (−0.141) (0.280) (0.232) (−0.059) (0.433) TA employees per thousands of people −0.154** −0.176*** −0.172** −0.163** −0.180*** −0.177*** (−2.350) (−2.601) (−2.496) (−2.488) (−2.644) (−2.577) Population (CA) −0.134* 0.027 0.103 −0.024 0.014 0.083 (−1.941) (0.169) (0.617) (−0.565) (0.101) (0.550) Municipality size 0.041** 0.046** −0.017 0.042** 0.046** −0.017 (2.043) (2.238) (−0.165) (2.072) (2.246) (−0.166) Municipality size squared −0.005** −0.006** −0.000 −0.005** −0.006** −0.000 (−2.134) (−2.357) (−0.027) (−2.166) (−2.370) (−0.026) Left −0.164*** −0.165*** −0.092 −0.164*** −0.165*** −0.092 (−11.245) (−11.307) (−1.248) (−11.258) (−11.302) (−1.248) Female 0.195*** 0.195*** 0.131* 0.195*** 0.195*** 0.131* (11.955) (11.974) (1.680) (11.962) (11.978) (1.680) Age 0.006** 0.006** −0.011 0.006** 0.006** −0.011 (2.171) (2.282) (−0.743) (2.173) (2.276) (−0.743) Age squared −0.000*** −0.000*** 0.000 −0.000*** −0.000*** 0.000 (−2.799) (−2.916) (0.678) (−2.803) (−2.910) (0.678) Head of household 0.010 0.010 0.017 0.010 0.010 0.017 (0.597) (0.595) (0.219) (0.593) (0.597) (0.219) Dummy married −0.017 −0.018 −0.123 −0.018 −0.018 −0.123 (−1.024) (−1.077) (−1.238) (−1.052) (−1.081) (−1.239) Dummy self-employed 0.108*** 0.106*** 0.078 0.108*** 0.107*** 0.078 (4.207) (4.149) (0.662) (4.223) (4.157) (0.662) Dummy retired 0.154*** 0.155*** 0.196 0.153*** 0.154*** 0.196 (5.463) (5.487) (1.342) (5.456) (5.484) (1.342) Dummy public employee −0.150*** −0.152*** −0.404*** −0.150*** −0.152*** −0.404*** (−7.476) (−7.557) (−3.944) (−7.479) (−7.554) (−3.945) Dummy low Unemployment risk (mean) −0.128*** −0.128*** −0.117 −0.128*** −0.128*** −0.117 (−6.132) (−6.110) (−1.148) (−6.128) (−6.108) (−1.148) Education level −0.017*** −0.017*** −0.020* −0.017*** −0.017*** −0.020* (−8.695) (−8.596) (−1.647) (−8.678) (−8.595) (−1.647) Nationalist −0.009 −0.008 −0.302* −0.010 −0.008 −0.302* (−0.272) (−0.237) (−1.683) (−0.280) (−0.243) (−1.683) Observations 30,891 30,891 30,891 30,891 30,891 30,891 Log-likelihood −22,641.947 −22,598.572 −22,339.425 −22,640.050 −22,598.391 −22,338.923 Fixed effects Yes No No Yes No No Time effects Yes Yes Yes Yes Yes Yes FE ×5 years TE No Yes Yes No Yes Yes Individual_Var.s × TE No No Yes No No Yes Note: z statistics are reported in parentheses and are derived using robust standard errors. *p < .10; **p < .05; ***p < .01. 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1891 TABLE 7 Falsification Test—The Determinants of Perceived Tax Pressure Along Time. Ordered-Probit, 1994–2015; Separate Regressions by UR Type (1) (2) (3) (4) Low UR High UR Low UR High UR GDP (CA) 0.257 −0.528 (0.916) (−1.475) Unemployment (CA) −0.042 −0.071 (−0.497) (−0.598) Deficit (CA) 3.680** −0.342 3.627** −0.174 (2.070) (−0.158) (2.038) (−0.080) FYL_%GVA_Secondary_sector 1.103 −3.496 1.343 −2.848 (0.590) (−1.245) (0.697) (−0.963) FYL_%GVA_Construction_sector −0.169 −1.808 −0.112 −1.433 (−0.063) (−0.471) (−0.042) (−0.372) FYL_% GVA_Tertiary_sector 0.722 −5.000 0.836 −4.216 (0.340) (−1.611) (0.387) (−1.332) Leftist government (CA) −0.072 0.078 −0.077 0.079 (−1.392) (1.118) (−1.474) (1.115) Electoral cycle (CA) −0.032 0.056 −0.035 0.065* (−1.185) (1.532) (−1.276) (1.811) TA employees per thousands of people −0.268*** −0.019 −0.269*** −0.024 (−3.200) (−0.159) (−3.204) (−0.204) Population (CA) −0.028 0.205 0.057 0.051 (−0.135) (0.783) (0.301) (0.216) Municipality size 0.077*** −0.010 0.077*** −0.009 (3.033) (−0.293) (3.036) (−0.268) Municipality size squared −0.009*** −0.000 −0.009*** −0.000 (−2.937) (−0.034) (−2.933) (−0.064) Left −0.189*** −0.127*** −0.189*** −0.127*** (−10.221) (−5.294) (−10.227) (−5.276) Female 0.212*** 0.171*** 0.212*** 0.170*** (10.219) (6.426) (10.221) (6.404) Age 0.004 0.005 0.004 0.005 (1.161) (0.926) (1.161) (0.914) Age squared −0.000* −0.000 −0.000* −0.000 (−1.771) (−1.072) (−1.773) (−1.061) Head of household −0.005 0.037 −0.005 0.037 (−0.230) (1.299) (−0.226) (1.301) Dummy married 0.018 −0.077*** 0.018 −0.077*** (0.804) (−2.979) (0.801) (−2.988) Dummy self-employed 0.135*** 0.010 0.135*** 0.010 (4.556) (0.182) (4.559) (0.174) Dummy retired 0.161*** 0.181** 0.161*** 0.179** (4.669) (2.261) (4.674) (2.232) Dummy public employee −0.141*** −0.176*** −0.141*** −0.176*** (−5.945) (−4.396) (−5.940) (−4.404) Education level −0.018*** −0.016*** −0.018*** −0.016*** (−7.017) (−5.100) (−7.011) (−5.099) Nationalist 0.013 −0.046 0.013 −0.045 (0.300) (−0.828) (0.299) (−0.808) Observations 19,074 11,817 19,074 11,817 Log-likelihood −14,130.416 −8,392.347 −14,130.725 −8,393.204 Fixed effects No No No No Time effects Yes Yes Yes Yes FE ×5 years TE Yes Yes Yes Yes Individual_Var.s × TE No No No No Note: z statistics are reported in parentheses and are derived using robust standard errors. *p < .10; **p < .05; ***p < .01. 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 1892 ECONOMIC INQUIRY Equations (1)–(3) by means of ordered logit administration is different from the Spanish one. and OLS models, respectively. Again, the results Finally, we think this line of research might merit qualitative results do not change. further theoretical developments. B. Falsification Test REFERENCES Table 6 presents the results obtained by per- forming a falsification test consisting in estimat- Alm, J., Y. Liu, and K. Zhang. “Financial Constraints and ing Equation (1) by substituting the perceived Firm Tax Evasion.” International Tax and Public Finance, 26(1), 2019, 71–102. tax enforcement for the perceived tax pressure Andreoni, J. “The Desirability of a Permanent Tax Amnesty.” as dependent variable. As expected, results sug- Journal of Public Economics, 45(2), 1991, 143–59. gest that tax policy does not react to AC-specific . “IRS as Loan Shark. Tax Compliance with Borrow- ing Constraints.” Journal of Public Economics, 49(1), economic cycle. Indeed, in the baseline models 1992, 35–46. (columns 1 and 4), where the filtering process is Auerbach, A., and Y. Gorodnichenko. “Measuring the Out- not completely taking into account all the com- put Responses to Fiscal Policy.” American Economic ponents of the dependent variable, the economic Journal—Economic Policy, 4(2), 2012, 1–27. Backus, P. G., and A. Esteller-Moré. “Risk Aversion and cycle seems to affect the tax policy as it is per- Inequity Aversion in Demand for Unemployment Bene- ceived by taxpayers in a procyclical way. This fits.” International Tax and Public Finance, 24(2), 2017, effect, though, vanishes as the filtering process is 198–220. Baretti, C. H., B. Huber, and B. K. Lichtblau. “A Tax on Tax fully implemented. Table 7 furthers this falsifica- Revenue: the Incentive Effects of Equalizing Transfers: tion test by allowing the effect of the economic Evidence from Germany.” International Tax and Public cycle to be different based on the individual risk Finance, 9(6), 2002, 631–49. type. Results fail to show a significant impact of Besley, T., and T. Persson. “The Origins of State Capacity: Property Rights, Taxation, and Politics.” American Eco- the economic cycle on tax policy both for high- nomic Review, 99(4), 2009, 1218–44. and low-risk individuals. These results validate Blank, J. D., and D. Z. Levin. “When Is Tax Enforcement the reliability of our identification strategy. Publicized?” Virginia Tax Review, 30(1), 2010, 1–38. Bönke, T., B. Jochimsen, and C. Schröder. “Fiscal Equaliza- tion and Tax Enforcement.” German Economic Review, 18(3), 2017, 377–409. V. CONCLUSIONS Brondolo, J. “Collecting Taxes During an Economic Crisis: Challenges and Policy Options.” IMF Staff Position Despite a strand of the literature on public Note, SNP/09/17, 2009. finance acknowledges tax enforcement is an addi- Chen, S. X. “The Effect of a Fiscal Squeeze on Tax Enforce- ment: Evidence from a Natural Experiment in China.” tional parameter of an optimal fiscal system (see Journal of Public Economics, 147, 2017, 62–76. e.g., Slemrod and Gillitzer 2014), there is lit- Esteller-Moré, A. “Is There a Connection Between the Tax tle literature checking whether this is the case. Administration and The Political Power?” International That is, there are not many positive analyses aim- Tax and Public Finance, 12(5), 2005, 639–63. . “Is the Tax Administration Just a Money Machine? ing at explaining how tax enforcement evolves Empirical Evidence on Redistributive Politics.” Eco- along the economic cycle. This is the challenge nomics of Governance, 12(3), 2011, 275–99. of this paper. Fishlow and Friedman. “Tax Evasion, Inflation and Stabiliza- tion.” Journal of Development Economics, 43(1), 1994, In particular, we estimate, first, if perceived 105–23. tax enforcement depends on the state of the econ- Ghosh, A. R., J. I. Kim, E. G. Mendoza, J. D. Ostry, and M. S. omy, and if so, second, estimate its nature (pro or Qureshi. “Fiscal Fatigue, Fiscal Space and Debt Sus- counter-cyclical). To do so, we have used survey tainability in Advanced Economies.” Economic Jour- nal, 123(566), 2013, F4–F30. data, as a proxy of real efforts by the tax admin- Gould, W. W. “sg19: Linear Splines and Piecewise Lin- istration and tried to filter any other potential ear Functions.” Stata Technical Bulletin, 15, 1993, (individual) explanation in the survey responses 13–17. Reprinted in Stata Technical Bulletin Reprints, that might bias our dependent variable. From the 3, 98–104. College Station, TX: Stata Press. Greene, W. H. Econometric Analysis. Pearson Education, analysis, we conclude that the perceived level of 2002. enforcement is cyclical with the nature of the International Monetary Fund. “World Economic Outlook, response dependent on the severity of the crisis. October 2010: Recovery.” Risk, and Rebalancing, 2010. Johnston, J., and J. DiNardo. Econometric Methods. 4th ed. These results confirm those of Chen (2017) while New York, NY: McGraw-Hill, 1997. extending those findings to subnational regions Kahneman, D., and A. Tversky. “Prospect Theory: An Anal- and severity of economic crisis. It would be inter- ysis of Decision Under Risk.” Econometrica, 47(2), esting to test this result using administrative data 1979, 263–91. Kuehn, Z. “Tax Rates, Governance, and the Informal Econ- on tax enforcement effort as well as in other con- omy in High-Income Countries.” Economic Enquiry, texts, where the institutional design of the tax 52(1), 2014, 405–30. 14657295, 2020, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ecin.12902 by Spanish Cochrane National Provision (Ministerio de Sanidad), Wiley Online Library on [07/11/2022]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License DURÁN-CABRÉ, ESTELLER-MORÉ & SALVADORI: CYCLICAL TAX ENFORCEMENT 1893 OECD (2011) Tax Administration in OECD and Selected Slemrod, J., and S. Yitzhaki. “The Optimal Size of a Tax Col- Non-OECD Countries: Comparative Information lection Agency.” Scandinavian Journal of Economics, Series (2010). 89(2), 1987, 183–92. . (2013) Tax Administration 2013—Comparative . “Tax Avoidance, Evasion, and Administration,” in Information on OECD and Other Advanced and Handbook of Public Economics, Vol. 3, edited by A. J. Emerging Economies. Auerbach and M. S. Feldstein. Elsevier, 2002, 1423–70. Pappa, E., R. Sajedi, and E. Vella. “Fiscal Consolidation with Sribney, W. M. “sg37: Orthogonal Polynomials.” Stata Tech- Tax Evasion and Corruption.” Journal of International nical Bulletin, 25, 1995, 17–18 Reprinted in Stata Tech- Economics, 96(1), 2015, S56–75. nical Bulletin Reprints, 5, 96–98. College Station, TX: Poirier, D. J., and S. G. Garber. “The Determinants of Stata Press. Aerospace Profit Rates 1951-1971.” Southern Eco- Vegh, C. A., and G. Vuletin. “How Is Tax Policy Conducted nomic Journal, 41(2), 1974, 228–38. over the Business Cycle?” American Economic Journal: Ramey, V. A., and S. Zuvairy. “Government Spending Mul- Economic Policy, 7(3), 2015, 327–70. tipliers in Good Times and in Bad: Evidence from US Wooldridge, J. Econometric Analysis of Cross Section and Historical Data.” Journal of Political Economy, 126(2), Panel Data. MIT Press, 2002. 2018, 850–901. Wren-Lewis, S. “Macroeconomic Policy in Light of the Richardson, G., G. Taylor, and R. Lanis. “The Impact of Credit Crunch: The Return of Counter-Cyclical Fis- Financial Distress on Corporate Tax Avoidance Span- cal Policy?” Oxford Review of Economic Policy, 26(1), ning the Global Financial Crisis: Evidence from Aus- 2010, 71–86. tralia.” Economic Modelling, 44(C), 2015, 44–53. Young, M., M. Reksulak, and W. F. Shughart II. “The Political Shaw, J., J. Slemrod, and J. Whiting. “Administration and Economy of the IRS.” Economics and Politics, 13(2), Compliance,” in Tax by Design: The Mirrlees Review, 2001, 201–20. edited by Institute for Fiscal Studies. New York: Oxford University Press, 2009. Slemrod, J. “Tax Compliance and Enforcement.” Journal of SUPPORTING INFORMATION Economic Literature, 57(4), 2019, 904–54. Slemrod, J., and C. Gillitzer. “Insights from a Tax-systems Additional supporting information may be found online Perspective.” CESifo Economic Studies, 60(1), 2014, in the Supporting Information section at the end of the article. 1–31. APPENDIX S1: ROBUSTNESS ANALYSIS