Introduction Appendix 1. Details on the estimation of the shadow economy... 3

Transcription

1

2 Table of contents Introduction... 2 Appendix 1. Details on the estimation of the shadow economy... 3 Appendix 2. omparison of the obtained estimates of the shadow economy with the estimates from other sources Appendix 3. Sectorial matching of the data used for the sectorial breakdown of the passive shadow economy Appendix 4. Estimation of the consumer payments structure in Appendix 5. Estimation of the impact of the proposed regulatory tools on the value of card and cash payments Appendix 6. Impact of the passive shadow economy on government revenues Appendix 7. Sensitivity analysis with respect to the share of cash vs. card payments in consumer transactions

3 Introduction This document is a part of the broader study that consists of (1) the report: Reducing the shadow economy through electronic payments (hereinafter referred to as the Report ), as well as (2) technical appendices (this document) and (3) individual country reports. 1 The Report analyses the shadow economy in the eight entral and Southern European countries and investigates the potential of different solutions to reduce the size of the non-observed economy. The current document comprises the technical appendices that provide more details on our approach to the measurement of the shadow economy and its breakdown, as well as to estimating the effects of various regulatory measures. 1 The Report, technical appendices and individual country reports are available on: 2

4 Appendix 1. Details on the estimation of the shadow economy In this appendix we discuss in a more detail the applied methodology for estimating the shadow economy and present additional results, which are not included in the Report. The structure of the appendix is determined by the distinguished main steps of the analysis, as described in hapter 2.1. of the Report. Step 1. urrency demand analysis (DA) Description of the approach and data used urrency in circulation (cash) conveys useful information about all (not only officially registered) economic activities. There are two kinds of demand for cash: Structural demand for cash reflecting the need for certain amount of cash to be used in normal economic activities (for instance, people hold cash for precautionary or liquidity purposes); Excessive demand for cash related to shadow market activities. In the estimation of the shadow economy based on the currency in circulation (currency demand analysis, DA), it is assumed that the share of an excessive cash component in the overall money demand (in the form of both cash and deposits) is equal to the share of the shadow market in the economy. The key advantage of this method is that it allows us to estimate the level of the shadow economy and its development over time with a very limited need to resort to external estimates. Another advantage of the currency demand approach, bearing in mind the context of this study, is its direct relation to the use of cash in the economy. However, the standard currency demand approach has also some drawbacks. First, it does not allow us to distinguish the committed and passive components of the shadow economy (see hapter 1 of the Report), which play a critical role in our analysis. Second, this method makes it difficult to investigate the impact of variables that may influence both the structural demand for cash (which is not related to the shadow economy) and its excessive part (related to the shadow economy). This issue applies, for example, to the electronic payments variables. Third, to distinguish the demand for cash driven by transactions in the official economy from the demand for cash related to the shadow economy transactions, a theoretical value from the money demand equation needs to be calculated (see below) under the assumption of no shadow economy. To address this issue, many authors adopt an unrealistic approach based, for example, on the tax rate set at the zero level, which may significantly bias the results 2. Inspired by the existing DA research 3 and aware of its weaknesses, we propose a modified approach that may be described in the following four substeps. 2 The zero tax approach (or zero incentives to move to shadow economy approach) was used by Tanzi V. The Underground Economy in the United States: Annual Estimates, , Staff Papers International Monetary Fund, Vol. 30, No. 2, 1983, pp For more recent examples, see Embaye A., "Underground economy estimates for non-oed countries using currency demand method, ," MPRA Paper 20308, University Library of Munich, Germany, 2007 and Ardizzi G., Petraglia., Piacenza M., Turati G., Measuring the underground economy with the urrency Demand Approach: A reinterpretation of the methodology, with an application to Italy, Review of Income and Wealth, vol. 60(4), 2014, pages This approach is also suggested by Ahumada H., Alvaredo F., anavese A. J., The Demand for urrency Approach and the Size of the Shadow Economy: A ritical Assessment, Berkeley Program in Law & Economics, Working Paper Series. 3 See Tanzi (1983), Ardizzi et al. (2014), op. cit. 3

5 First substep The first substep consists in an econometric estimation of the currency demand equation: ( cash in circulation M1 ) i,t = α i + β 1 x 1,i,t + β 2 x 2,i,t + β 3 x 3,i,t, (A1.1) where i represents the analysed country and t stands for the analysed time period (in our estimation, we use quarterly data). In this equation, the explained (dependent) variable is the share of currency in circulation ( cash ) in the M1 monetary aggregate ( total transactional money 4 ). To explain its variation, we use the following explanatory variables 5 : Typical shadow economy determinants (x 1 ). The impact of these variables on the dependent variable is mostly related to changes in the size of the shadow economy. These relations are in line with the economic theory and confirmed by the empirical research. Variables used: the level of taxes and social security contributions, the rule of law index, unemployment rate. Payment card system variables (x 2 ). Higher levels of these variables (reflecting better development of the payment card system) may be associated with the two effects: (1) a decrease in the size of the shadow economy (by discouraging/hindering transactions in the shadow economy) and (2) replacement of the registered cash transactions with electronic payments (no impact on the size of the shadow economy). Variables used: number of payment cards per capita, ratio of the number of payment terminals to the number of payment cards 6 7. Other control variables (x 3 ). These variables, after controlling for the influence of x 1 and x 2 determinants, should not (directly) impact on the size of the shadow economy, but may still have some influence on the dependent variable. These variables are related to the level of the economic development, technical progress and institutional factors (some of which might be unobservable and must therefore take the form of dummies). Variables used: GDP per capita in PPS, the share of the population with Internet access 8, binary variables controlling for the changes in the dependent variable related to the euro adoption process in Slovakia and Slovenia, country dummies (fixed effects, discussed below). β 1, β 2 and β 3 represent vectors of the corresponding coefficients. Description of the variables used and their sources is presented in Table A1.1. In our analysis, we use the data for the variables discussed above for the following countries: Analysed countries: Bosnia and Herzegovina, Bulgaria, roatia, the zech Republic, Poland, Serbia, Slovakia, and Slovenia (countries that we focus on in the Report); 4 M1 monetary aggregate ( total transactional money ) includes cash and demand deposits held in financial institutions. 5 The explanatory variables have been chosen based on the literature review and statistical significance of the variables. 6 We use the number of payment cards per capita and the number of payment terminals per payment card instead of payment cards transaction value, because the former variables seem to be more exogenous (i.e. there are less feedback effects from the explained variable to the explanatory variables that are detrimental to the quality of estimation). 7 We use the number of payment terminals per payment card instead of the number of payment terminals per capita to avoid the collinearity of this variable with the number of payment cards per capita (collinearity occurs when explanatory variables are correlated with each other and hence produce imprecise estimates of parameters). 8 It may be argued that GDP per capita and/or the share of the population with internet access also influence the size of the shadow economy. However, the three most likely channels of this influence are controlled for in our model by other variables: (1) the rule of law/quality of public institutions (we control for it through the use of the World Bank s rule of law index), (2) changes in the business cycle (we control for it with the unemployment rate variable), (3) technological progress in the payment methods (we control for it with variables related to payment cards). Still, GDP per capita and/or the share of the population with internet access variables have turned out to be significant and robust in explaining some differences and fluctuations in the dependent variable. Therefore, we have kept them in our model and treated them as control variables. 4

6 Auxiliary countries: Denmark, Hungary, Norway, Sweden, and the United Kingdom (countries with their own currencies, which have been included in the econometric analysis to improve the quality of our estimates) 9. Table A1.1. Variables used in the currency demand analysis. Name of the variable Description of the variable Source(s) Explained (dependent) variable MONEY I. Typical shadow economy determinants TAXES_SOIAL RULE_OF_LAW UNEMP II. Payment card system variables ARDS_PER_APITA TERMINALS_NO/ARDS_NO III. Other control variables GDP_REAL_PPS_PER_APITA INTERNET_AESS SLOVAKIA_2009 SLOVAKIA_TRANSITION1 SLOVAKIA_TRANSITION2 SLOVENIA_2007 SLOVENIA_TRANSITION2 Source: EY The share of the currency in circulation in the sum of the currency in circulation and deposits held in financial institutions (M1 monetary aggregate), % of M1, seasonally adjusted Taxes and social security contributions, % of GDP, seasonally adjusted, quarterly observations (for Bosnia and Herzegovina interpolated from annual data) The value of the indicator measuring the rule of law from the Worldwide Governance Indicators; ranges from approximately -2.5 (weak rule of law) to 2.5 (strong rule of law), quarterly observations interpolated from annual data Unemployment rate, % of total labour force (economically active population), seasonally adjusted The number of payment cards per capita, seasonally adjusted, quarterly observations interpolated from annual data The ratio of the number of payment terminals to the number of payment cards, quarterly observations interpolated from annual data Real GDP per capita, EUR PPS in constant 2010 prices, seasonally adjusted (for Bosnia and Herzegovina quarterly observations interpolated from annual data) The share of the population with Internet access, % of population, quarterly observations interpolated from annual data Binary variable controlling for euro adoption in Slovakia in 2009 Binary variable controlling for the changes in the monetary aggregates in Slovakia 2 quarters before euro adoption Binary variable controlling for the changes in the monetary aggregates in Slovakia 1 quarter before euro adoption Binary variable controlling for euro adoption in Slovenia in 2007 Binary variable controlling for the changes in the monetary aggregates in Slovenia 1 quarter before euro adoption International Monetary Fund, central bank databases Eurostat, EcoWin (ministries of finance, central statistical offices) World Bank Worldwide Governance Indicators Eurostat, EcoWin (World Bank) European entral Bank, Masterard, National Bank of Serbia, entral Bank of Bosnia and Herzegovina European entral Bank, Masterard, National Bank of Serbia, entral Bank of Bosnia and Herzegovina Eurostat, World Bank International Telecommunication Union (United Nations) 9 It is controversial and complicated to include the developed euro area member states in the analysis due to the lack of reliable data on the currency in circulation in these countries. Some approximations are, however, possible for the countries within our sample that adopted the euro only recently, i.e. Slovenia and Slovakia. 5

7 The sample covers the available quarterly observations for these countries over the years of To estimate the equation (A1.1), we use the panel econometric methods, which allow us to conduct the analysis of multiple countries observed over multiple years. In particular, we apply the so called fixed effects panel estimator. This allows us to include in the estimation those time-invariant, unobservable characteristics of a given country (α i in equation (A1.1)) that affect the demand for cash in this country (the scale of this effect is different in each country). This is achieved by adding a separate dummy variable for each country. Panel data makes it possible to exploit this kind of effects because we observe each country in more than one time period. To illustrate the idea of fixed effects estimation, consider a hypothetical situation in which we have a single-period dataset at our disposal. It would then be impossible to say whether the demand for cash in one country is high because of, for example, a low number of cards per capita, or because of some unobservable cultural factors (which may be assumed to be constant over time, at least in a few-year period). In contrast, if we have two or more time periods in our sample, we can observe changes in cash demand along with changes in the number of cards per capita, while the unobservable cultural factors are kept constant. To split the estimated coefficients for the payment card system variables (x 2 ) into the two already described effects (i.e. related and unrelated to the shadow economy), we assume that the proportion of these effects is analogous to the proportion of the average impact on the level of the dependent variable of: typical shadow economy determinants (x 1 ) and other control variables (x 3 ), excluding dummy variables. We adopt this assumption for the sake of a conservative approach towards the impact of electronic payments on the size of the shadow economy. By contrast, an assumption that the whole impact of the payment card system variables is related to the shadow economy would most likely result in the overestimation of the positive effects of non-cash payments 11. Second substep In the second substep, we set the values of x 1 and x 2 vectors in equation (1) at their best observable levels for European countries (e.g. the lowest tax and social security contributions, the highest number of payment cards, etc.) and estimate the theoretical value of the explained variable. 12 Once again we adopt a conservative approach here, in contrast to the literature where it is common to set some of the discussed best values, especially for the tax-related variables, at the zero level (even though the zero tax burden does not exist in any economy and, therefore, is purely a theoretical concept) 13. For the dependent variable, the difference between the fitted value, calculated on the basis of the actual values of x 1 and x 2, and the estimated best theoretical value may be interpreted as the share of cash in the M1 aggregate that is related to shadow economy transactions. Given the observed stock of the M1 aggregate for a given country and period, the obtained difference allows us to calculate the amount of cash that is attributable to the shadow economy. 10 The available data covers different time periods for different countries. While it includes the whole time period for Slovakia, Slovenia and three of the auxiliary countries, it covers only the time period for Serbia (the shortest available sample in the group of the analysed and auxiliary countries). Variables that have not been available at a quarterly frequency have been interpolated into quarterly data. 11 The exact influence of the payment system variables on the shadow economy (in particular on the passive shadow economy) is examined with the use of the MIMI model in step 3 of the analysis. 12 For the payment card system variables we use the adjusted coefficients so that they are related only to changes in the shadow economy level. 13 The zero tax approach (or zero incentives to move to shadow economy approach) was used by Tanzi (1983), op. cit. For more recent examples, see Embaye A. (2007), "Underground economy estimates for non-oed countries using currency demand method, ," MPRA Paper 20308, University Library of Munich, Germany, and Ardizzi et al. (2014), op. cit. This approach is also suggested by Ahumada H., Alvaredo F., anavese A. J. (2004), The Demand for urrency Approach and the Size of the Shadow Economy: A ritical Assessment, Berkeley Program in Law & Economics, Working Paper Series. 6

8 Third substep In the third substep, we estimate the size of the shadow economy 14 with the use of the quantity theory of money. In particular, we assume that the ratio of: (a) the total output (including the shadow economy) to (b) the M1 monetary aggregate is equal to the ratio of: (c) the output in the shadow economy to (d) the estimated amount of cash attributable to the shadow economy. In other words, the velocity of money in the shadow economy, (c)/(d), is assumed to be equal to the velocity of money in the overall economy, (a)/(b). Transforming the obtained equation and knowing the values of (b) and (d), we may estimate the share of the shadow economy output in the total output (including also the shadow economy) without knowing the exact value of the velocity of money. The approach presented here seems superior to the approach that is often applied in the literature and which consists in translating the amount of cash attributable to the shadow economy (obtained from the currency demand analysis) into the value of the shadow economy output with the use of some specific estimate of the velocity of money (sometimes based on dubious and/or outdated figures). The problem arises because, in order to calculate the exact value of the velocity of money, one has to know the part of the GDP that is not related to shadow economy activities. In particular, the velocity of money in the whole economy should not be assumed to be equal to the ratio of the M1 monetary aggregate to the sum of the official GDP figure and the size of the shadow economy estimated with the DA such an approach includes a double counting of at least some part of the shadow economy, since the official GDP figures already include statistical offices estimates of the shadow economy (which may be higher or lower than the estimates obtained with the DA). One solution to the above challenge is to simply ignore the fact that the official GDP figures already include the estimated shadow economy, as in Tanzi s approach 15. Another solution, described by Ardizzi et al. 16, Warner et. al. 17 and partly mentioned by Schneider et al. 18, is to assume that in some base period the share of the shadow economy in the official GDP equals zero, and the velocity of money is obtained by dividing the value of GDP in that period by the legal component of M1 (obtained after subtracting the shadow money value that was estimated with the use of the DA). Assuming that the velocity of money is constant over time, one may then calculate the monetary level of the shadow economy in all the periods of the analysis. However, we find the constant velocity assumption highly implausible. In our approach, we do not have to make either of the above assumptions. We take advantage of the OED publication discussing the shadow-economy components within the official GDP estimates in various OED countries in the years 2005 and We also refer to the national statistical offices in order to obtain data on the shadow economy component included in the official GDP figures for the years other than mentioned above and for the non-oed countries (such as Serbia or Bosnia and Herzegovina). This allows us to calculate the registered GDP volumes and thereby obtain the time-varying velocity of money estimates. Therefore, in our opinion, the estimates of the shadow economy presented in this Report, expressed as % of official GDP 21, should be more reliable than the estimates based on the approaches described above, requiring the adoption of additional, implausible assumptions. 14 The size of the shadow economy corresponds to the part of output / GDP that is generated in the shadow economy. 15 Tanzi (1983), op. cit. 16 Ardizzi et al. (2014), op cit. 17 Kyle S.., Warner, A., Dimitrov L., Krustev R., Alexandrova S., Stanchev K. (2001), "Measuring the Shadow Economy in Bulgaria," Working Papers , ornell University, Department of Applied Economics and Management. 18 Schneider F., Buehn A., Montenegro. E. (2010), "Shadow Economies All over the World: New Estimates for 162 ountries from 1999 to 2007," Policy Research Working Paper, 5356, The World Bank. 19 OED (2014), The Non-Oberved Economy in the System of National Accounts, Statistics Brief, No. 18, The estimate of the shadow economy that is added to the official GDP figure is usually not directly reported by the statistical offices. 21 We present our estimates in this form to make them more comparable to other estimates quoted in the literature (even though some of them are affected by the described methodological errors). 7

9 Fourth substep Having obtained the initial estimates of the shadow economy, we additionally account for the fact that even in a country with the best values in x 1 and x 2 vectors (at the level of the best performing countries in a given area), the shadow economy would not disappear completely. In other words, there would still be some low, say, natural level of the shadow economy (e.g. some illegal transactions will not begin to be reported because of lower taxes and more payment cards). We estimate this natural level of the shadow economy as an average of the four lowest levels of the shadow economy measured by statistical offices in OED countries 22. By adding the above calculated average (equal to 1.95% of the official GDP) to the initially obtained estimates of the shadow economy, we arrive at the final estimates of the overall shadow economy. To sum up, our contribution to the literature on the DA and shadow economy estimation comprises: the inclusion of the variables related to the payment card system, avoiding some controversial assumptions as regards the velocity of money, and calibration of the lowest possible level of the shadow economy instead of adopting the implausible assumption of no shadow economy in a non-existent zero tax economy. Estimation results Table A1.2 presents the results of our econometric estimation of the currency demand equation. The estimates of the overall level of the shadow economy, based on the described methodology and the results included in Table A1.2, are discussed in hapter 2.3 of the Report and individual country reports. It is worth noting that the shadow economy figures for 2014 are based on the estimates/forecasts of some of the shadow economy determinants. It applies to: (1) the rule of law index (assumed to be equal to its value in 2013, since this variable changes very slowly in various countries) and (2) the payment card system variables (estimates based on the data obtained from Masterard). 22 These countries are: Norway, anada, the Netherlands, and the United Kingdom (the Netherlands and the United Kingdom had the same shadow economy level according to the statistical offices, so we used the average of 4 countries, instead of 3). For these countries, respective values of x 1 and x 2 vectors from equation (A1.1) are close to the best observable levels of x 1 and x 2 variables. The data comes from the OED (2014), op. cit. The estimates of statistical offices are based, among other things, on the national accounts data, the labour market data and on special consumer surveys. The approaches adopted by statistical offices differ from our approach, since, for example, they are not based on econometrics. 8

10 Table A1.2. Econometric estimates of the currency demand equation. MONEY Explained (dependent) variable I. Typical shadow economy determinants Parameter Std. Dev. p-value TAXES_SOIAL ** RULE_OF_LAW *** UNEMP *** II. Payment card system variables Parameter Adjusted parameter Std. Dev. p-value ARDS_PER_APITA *** TERMINALS_NO/ARDS_NO ** III. Other control variables Parameter Std. Dev. p-value GDP_REAL_PPS_PER_APITA *** INTERNET_AESS *** R-squared Mean dependent variable Adjusted R-squared Std. dev. dependent variable Std. Error of regression Sum squared residuals Durbin-Watson statistic F-statistic Periods included 56 Number of countries 13 Total panel (unbalanced) observations 537 Statistical significance levels *10%, **5%, ***1%. Notes: Estimates for the binary control variables are not presented here. For the payment card system variables, adjusted parameters describe their impact only on the shadow-economy-related money demand. Source: EY Step 2. Labour market analysis (LMA) To decompose the overall volume of the shadow economy into the committed and passive segments, we compare the two different measures of employment: data on the official employment and employment data from the Labour Force Surveys. The former is officially reported, while the latter is based on declarations in the survey questionnaires and aims at measuring the total employment, including activities in the shadow economy. We interpret the gap between the declarative and official employment as a proxy of the committed shadow economy activities. This is based on the assumption that the output of the committed shadow economy is correlated with and mirrored by shadow labour force inputs 23. It should be noted, however, that unreported employment is also possible in the registered companies that are not involved in the committed shadow economy (see hart 1.1 in the Report). Therefore, from this perspective, our assumption can result in an overestimation of the committed and an underestimation of the passive component. On the other hand, however, we do not account for the 23 It is worth to recall that the passive shadow economy consists in underreporting of the revenues by registered, legally operating entities. 9

11 fact that some companies with no unreported labour force may also be involved in the committed shadow economy. Before we use the obtained employment gap to estimate the committed shadow economy production, we account for the two additional factors suggesting that the gap between the declarative and official employment is higher than the respective production gap. First, we account for the fact that the number of hours worked by an officially employed person is, on average, higher than in the case of a person working in the shadow economy. To estimate this difference, we use the data on working hours for both the official and (calculated) shadow economy employment, on the basis of Eurostat (LFS) data and data on working hours in the non-observed employment provided in the dedicated research 24. Second, we take into consideration the fact that the average labour productivity of officially employed people is higher than of those working in the shadow economy. As a measure of the relative productivity in the official vs. shadow economy we use the ratio of the official average monthly wage to the minimum wage. Such an approach is essentially an approximation of the actual labour productivity as the labour productivity does not necessarily correspond to the level of remuneration 25. In addition, for the shadow economy, only estimates of the average remuneration are available and it remains unclear whether this remuneration is lower or higher than the official minimum wage. However, the evidence available for Poland suggests that in 2010 the average fulltime equivalent remuneration in the shadow economy was slightly lower than the official gross minimum wage 26. This would imply that our approach to estimating the productivity of unregistered workers may result in overestimation of the committed and the respective underestimation of the passive component of the shadow economy. This, in turn, would be consistent with our preference to be on the conservative side rather than presenting biased, overestimated figures that could weaken the credibility of our conclusions. After applying the two above-discussed corrections, the adjusted employment gap allows us to approximate the share of the committed shadow economy in the whole economy. To obtain the estimates of the passive component, we simply subtract the estimates of the committed shadow economy from the estimated overall level of the shadow economy. We refer to the thus obtained output of the labour market analysis (LMA) as to the base estimates. Step 3. MIMI model Description of the approach and data used The outcome of steps 1 and 2 of the analysis are: the estimates of the overall level of the shadow economy, the quantified impact of its determinants, and the base estimates of the passive shadow economy. Improving the quality of the passive shadow economy estimates and identification of the key determinants of this component of unregistered activities are the goals of step 3 of the analysis. For this purpose, we apply the Multiple Indicators Multiple auses Model (MIMI). In the MIMI model, the (passive) shadow economy is an unobservable variable. However, changes in this variable over time can be measured through the identification and the analysis of two kinds of variables (see hart A1.1): auses that explain the size of the unobserved variable; Indicators that result from the unobserved variable. 24 European ommission (2007), Undeclared Work in the European Union, Eurobarometer. 25 In addition, according to the survey conducted by the Lithuanian Free Market Institute in the Baltic Sea Region (including Poland), a considerable share of employees on legal job contracts receive part of their wage in the form of an envelope wage (undeclared wage). See Vytautas Zukauskas (editor, 2015), Shadow Economies in the Baltic Sea Region 2015, Lithuanian Free Market Institute. [online, accessed ]. 26 entral Statistical Office in Poland (2011), Unregistered Employment in Poland in 2010, Statistical Information and Elaborations, [online, accessed ]. 10

12 To achieve the reliable results, in our model the only factor through which the causal variables influence the indicators should be the passive shadow economy. hart A1.1. Relation between the causes, shadow economy and indicators in the MIMI model. Source: EY The MIMI approach has important advantages. It allows us to use a broad set of causes and indicators of the non-observed economy, whose selection should be determined by the analysed profile of the shadow economy. In our analysis, the MIMI model allows us to focus on the passive component of the non-observed economy. It also takes into account a multitude of variables, and hence reduces the risk of measurement errors. In our MIMI model, we use quarterly data available for the years for the eight analysed countries: Bosnia and Herzegovina, Bulgaria, roatia, the zech Republic, Poland, Serbia, Slovakia, and Slovenia. Since we focus on the passive shadow economy, we use the following variables in the model: auses of the passive shadow economy: o Total value of payment card transactions at physical terminals per capita, as % of GDP; o Rule of law index; o Total value of taxes and social security contributions, as % of GDP; Indicators of the passive shadow economy: o Base estimates of the passive shadow economy from step 2 of our analysis; o The difference between the standard rate of VAT and the ratio of the actual VAT revenues to the domestic demand 27 ; o Final electricity consumption per 1 unit of the real gross domestic product. For a more detailed description of these variables and their sources see Table A1.3. These variables have been chosen based on: our understanding of the passive shadow economy nature, available data and statistical reasoning. Other variables, especially the potential causes of the passive shadow economy, have also been tested, but could not be included in our estimation for a number of reasons. For instance, some variables that we wanted to use in the MIMI model turned out to be collinear with the variables already included in the model or were statistically insignificant. An important variable that was not included in the MIMI model is GDP per capita, which was collinear with the card transaction value (as % of GDP). It should be noted that the card transaction value (as % of GDP) is directly related to the passive shadow economy, whereas GDP per capita is related also to the committed component. Therefore, since we use the MIMI model to estimate the evolution of the passive shadow economy, we should rather opt for the use of the card transaction value (as % of GDP). In addition, it should be emphasised that the decision not to use the GDP per capita variable in the MIMI model does not (significantly) influence the obtained results with 27 Here, for simplicity, we use the domestic demand (private consumption plus gross fixed capital formation plus government expenditures) as a proxy of the VAT base. However, in the estimation procedure of the impact of the passive shadow economy contraction on government revenues, we use a more detailed approach to calculate the VAT base. 11

13 respect to the impact of particular regulations on the passive shadow economy and government revenues. The total value of payment card transactions (in % of GDP) is a natural candidate for a cause of the passive shadow economy, which is driven by retail cash transactions. We calculate the card transactions to GDP ratio in order to account for the changes in the popularity of card vs. cash payments in the economy 28. The other two causes included in our MIMI model (the rule of law index and the taxation burden) have been proven to be relevant determinants of the overall level of the shadow economy (in step 1 of our analysis), while not being specific to the committed component only. Therefore, there is no reason to assume that they do not affect the passive shadow economy. Obviously, the base estimates of the passive component, obtained in the previous steps of our analysis, should be a good indicator variable. Since the passive shadow economy is related to unreported consumer transactions, a growing difference between: (1) the standard VAT rate and (2) the ratio of actual VAT revenues to the domestic demand may suggest that the passive shadow economy is increasing 29. Moreover, an increasing divergence between the final electricity consumption (electricity is used for the whole production, including the passive shadow economy) and real gross domestic product (official statistics may not comprise the whole passive shadow economy 30 ) may indicate that the passive non-observed economy is growing Here we assume that the ratio of the total transaction value to GDP is, more or less, stable. 29 However, a growing difference between: (1) the standard VAT rate and (2) the ratio of actual VAT revenues to the domestic demand may also reflect changes that are not related to the evolution of the passive shadow economy (e.g. a growing number of missing trader frauds). 30 One should remember that official GDP figures include some estimates of the shadow economy delivered by statistical offices (these estimates, however, are rarely available to public - OED (2014), op. cit. is a unique publication providing such data, however only for a limited number of OED countries in a single year 2012). 31 However, a growing divergence between the final electricity consumption and real gross domestic product may also reflect changes that are not related to the evolution of the passive shadow economy (e.g. technological progress that lowers the energy use in a given industry or an increased use of the electricity by entities active in the committed shadow economy). 12

14 Table A1.3. Variables used in the MIMI model. Name of the variable Description of the variable Source(s) auses of the passive shadow economy log_transations_value_ TO_GDP RULE_OF_LAW TAXES_SOIAL Indicators of the passive shadow economy DA_LM_PASSIVE_ESTIMATE Logarithm of the ratio of the total value of payment card transactions at physical terminals to GDP, for some time periods and countries we have applied quarterly interpolation from annual data The value of the indicator measuring the rule of law from the Worldwide Governance Indicators, ranges from approximately -2.5 (weak rule of law) to 2.5 (strong rule of law), quarterly interpolated from annual data Taxes and social security contributions, % of GDP, seasonally adjusted (quarterly data interpolated from annual data for Bosnia and Herzegovina) The estimated size of the passive shadow economy based on the estimated overall level of the shadow economy from the DA model and labour market analysis European entral Bank, Masterard, Eurostat, National Bank of Serbia World Bank Worldwide Governance Indicators Eurostat, EcoWin (ministries of finance, central statistical offices) EY calculations (see steps 1 and 2) VAT_DIFF The difference between the standard VAT rate and the ratio of the actual VAT revenues to the domestic demand EY calculations based on Eurostat, OED and additional data sources, including national ministries of finance ELETRO_SHARE Source: EY Final electricity consumption per 1 unit of the real gross domestic product, in EUR Purchasing Power Standards in constant 2014 prices, seasonally adjusted, quarterly interpolated from annual data Eurostat, World Bank Despite its advantages, the MIMI approach has also some drawbacks. In particular, the MIMI estimation generates only an index showing deviations from some long-run average size of the shadow economy, in contrast to the level of the shadow economy. To obtain the level of the shadow economy, a reference to an external estimate of the shadow market must be used. In some of the studies available in the literature, the sources of these external estimates are not clear, and some authors seem to rely on outdated estimates or expert guesses. In addition, in the literature such external estimates are often taken for a single year, which means that the obtained results may be biased by some random, one-off fluctuations in the reference period. After transforming the output of the MIMI model, such random fluctuations in a single year might significantly affect the entire estimated path of the shadow economy. In order to address this issue, we use a standard deviation and the sample (long-run) averages of our own estimates of the passive shadow economy. As a result, in our approach, the impact of one-off fluctuations influencing the shadow economy estimates for single years is averaged out, which should allow us to obtain more reliable estimates of the path of the shadow economy. Another shortcoming of the MIMI approach, as evidenced in the literature, is that its results are heavily influenced by a discretionary decision of researchers. Since the output of the MIMI model (in the form of deviations from the average) takes both positive and negative values, it needs to be adjusted; otherwise, the level of the shadow economy would be negative (!) in some periods of time. To deal with this issue, some researchers 32 add an arbitrary constant to the MIMI output, calculate a new index and multiply it by the initial level of the shadow economy. Unfortunately, the selection 32 See, for example, Schneider, Buehn and Montenegro (2010), op. cit. 13

15 of different levels of the arbitrary constant may significantly affect the estimated level and dynamics of the shadow economy (see Appendix 2 for more details and examples). On the one hand, the estimation procedure for our MIMI model does not significantly differ from the approach described in the literature 33. On the other hand, however, we address the weaknesses outlined above through an innovative approach to transformation of the output of the MIMI model into the final estimate of the shadow economy (here, the passive shadow economy). The way we achieve this is as follows. First, we standardise the output of the MIMI model in order to obtain a variable that describes the deviations from some long-run average level of the passive shadow economy. Next, we conduct a reverse standardization of this variable, which includes the following steps: We multiply the standardised MIMI latent variable with the standard deviation of the base estimate of the passive shadow economy, obtained from step 2 of our analysis. We calculate a single standard deviation for all the countries. In order to obtain it, we must first calculate the deviations of the passive shadow economy from the country-specific, long run averages. This transformation makes our results consistent with the fixed effects panel approach that we use in our econometric models. Next, we pool the deviations of the passive shadow economy from the long-run country specific averages and calculate a single estimate of the shadow economy standard deviation; We add the country averages (the long-run averages) of the base estimate of the passive shadow economy from step 2 of our analysis. This way, we make the level and the scale of variation of the final passive shadow economy estimates correspond to the level and the scale of variation of the base estimates. Therefore, the MIMI model provides us with an additional information on the evolution of the passive shadow economy over time. The presented approach is our innovative contribution to the literature and makes the process of transforming the MIMI output more formalised and tractable. Importantly, our approach allows us to interpret the estimated coefficients of the causes variables in terms of their impact on the passive shadow economy. By contrast, in the literature that we are aware of, the estimated coefficients of the MIMI models have rather a technical meaning and therefore do not allow the authors to interpret them in terms of the magnitude of their impact on, say, the size of the shadow economy. In addition, in our MIMI model we account for the marginal decreasing influence of electronic payments on the size of the passive shadow economy by introducing into the model the logarithm of non-cash payments to GDP ratio (this form of the variable better fitted the data). Estimation results Table A1.4 presents the results of the econometric estimation of the MIM model. The estimates of the passive shadow economy, based on the described methodology and the results included in Table A1.4, are discussed in hapter 2.3 of the Report and individual country reports. It is worth noting that the passive (and committed) shadow economy figures for 2014 are based on the estimates/forecasts of some of the shadow economy determinants. It applies for: (1) the rule of law index (assumed to be equal to its value in 2013, since this variable changes very slowly in various countries), (2) the total value of payment card transactions at physical terminals per capita (estimates based on the data obtained from Masterard). 33 For more details on the MIMI estimation procedure see, for example, Schneider F., Buehn A. (2013), "Estimating the Size of the Shadow Economy: Methods, Problems and Open Questions,", ESifo Working Paper Series, 4448, ESifo Group Munich. 14

16 Table A1.4. Econometric estimates of the MIM model in which the latent variable is the passive shadow economy. auses Parameter Standardised Empirical p-value log_transations_value_ TO_GDP *** RULE_OF_LAW ** TAXES_SOIAL *** Indicators Parameter Standardised Empirical p-value DA_LM_PASSIVE_ESTIMATE NA VAT_DIFF *** ELETRO_SHARE *** Periods included 52 Number of countries 8 Total panel (unbalanced) observations 268 Statistical significance levels *10%, **5%, ***1%. Notes: Parameter denotes the original parameters from the MIMI. Standardised parameters show the relation between an increase in a given variable by one standard deviation and the change in the passive shadow economy expressed in its standard deviations. Empirical parameters present the relation between an increase in a given variable by one unit and a change in the passive shadow economy expressed in percentage points of official GDP. Source: EY In the context of our analysis, it is more accurate to interpret and use the results of the MIMI rather than the DA model. The MIMI model allows us to better capture the evolution of the passive shadow economy over time and to obtain more precise estimates of the impact of various variables on this component of the shadow economy in the analysed countries. This is because the variables used in the MIMI model are closely related to the passive shadow economy, whereas the variables used in the DA approach are supposed to capture the developments of the overall shadow economy, including the committed component. Step 4. Sectorial Structure Analysis Description of the approach and data used In the final step, we decompose the estimated passive shadow economy volume into the different sectors of the economy. Due to limited data availability, the estimation of the sectorial breakdown requires complex transformations with the use of data from different sources, in particular: Data on the level of final consumption expenditure of households (source: Eurostat); Data on the sectorial breakdown of consumption expenditure of households (including nonresidents, mainly tourists), used in calculations of the Harmonised Index of onsumer Prices (HIP, source: Eurostat); Data on the sectorial breakdown of the value of Masterard s payment card transactions at physical terminals (source: Masterard); Data on the Masterard s market shares (source: Masterard). Since some of the required data is available only for single time periods, which additionally differ among the analysed countries, we interpret the estimated shares of individual sectors in the passive shadow economy as their long-term averages. Necessary data adjustments and subsequent calculations may be divided into the six substeps, conducted separately for each of the analysed countries. 15

17 First substep In the beginning, we harmonise and unify the sectorial breakdown of the data obtained from the sources described above. This transformation is necessary, because Masterard uses different sectorial classification than the statistical offices (for details, see Appendix 3). Second substep Next, we assume that the passive component is equal to zero in certain sectors for which the passive shadow economy transactions are unlikely. These sectors comprise: Electric appliances, telephone equipment, travel agencies and insurance (consumers have a high motivation to ask for a receipt in these sectors, not least for the sake of warranty or ability to execute their contract); Utilities and telephone services (sometimes these services are provided by the entities owned or controlled by the government; in addition, in these sectors, the calculation of the amount due for the service and its payment are very often separated in time, and payment is made against an invoice issued by the service provider); Housing maintenance, including services (the shadow economy in this sector is mostly of the committed nature). Third substep For each country, we calculate the total amount of households monetary spending on consumption by subtracting from the value of final consumption of households the value of imputed rents (which are not monetary), based on Eurostat data. Next, for each sector we multiply the obtained value by its corresponding HIP weight. 34 These weights, provided by Eurostat, capture the sectorial structure of households consumption in a given country (comprising also the expenditures of nonresidents, mainly tourists). The obtained products are our estimates of the value of households monetary spending on consumption in a given sector. Fourth substep Based on data for: (1) the value of Masterard s payment card transactions at physical terminals in different sectors and (2) the Masterard s market shares in the total number of payment cards in various countries, we estimate the total market value of payment card transactions at physical terminals in different sectors. Here we assume that the sectorial structure of the whole payment card market is analogous to the structure of the Masterard s payment card transactions at physical terminals. Fifth substep The passive shadow economy is mostly related to households consumption spending. However, we have data available on the total value of card payments, including also transactions made by companies. Therefore, we need to distinguish that part of card transactions, which are performed exclusively by consumers in different sectors. 34 Since for Bosnia and Herzegovina and Serbia HIP weights are not available, we assume that they are equal to weights for Bulgaria the country with the most comparable level of GDP per capita and geographic location in our sample. 16