## wp1817 — 2. THEORETICAL CONSIDERATIONS ## Source details **Canonical URL:** [wp1817 — 2. THEORETICAL CONSIDERATIONS](https://www.imf.org/-/media/files/publications/wp/2018/wp1817.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2018/wp1817.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2018/wp1817.pdf.json) --- ### Conceptual framework and structural equation - Shadow economy (SE) participation decision: - SE is negatively related to the probability of detection p and potential fines f. - SE is positively related to opportunity costs of remaining formal B, which are positively determined by taxation burden T and labor costs W. - Probability of detection p depends on enforcement actions A and facilitating activities F undertaken by individuals to reduce detection. - Structural formulation (variables as in source): SE depends on p, A, F, f, B, T, W in the form indicated in the source text. ### Definition and scope of the shadow economy - Shadow economic activities (SE) circumvent government regulation, taxation or observation. - Narrow definition: monetary and non-monetary transactions of a legal nature that would generally be taxable if reported. - Activities deliberately concealed to avoid: - payment of income, value added or other taxes and social security contributions; - compliance with legal labor market standards (minimum wages, maximum working hours, safety standards) and administrative procedures. - Focus: productive activities normally included in national accounts but remaining underground due to tax or regulatory burdens. - Exclusion: informal household activities (do-it-yourself activities and neighborly help) are typically excluded. ### Causes and indicators of informality (summary) - Determinants highlighted (as in Table 1): - Tax and social security contribution burdens. - Quality of institutions and corruption. - Regulations and enforcement intensity. - Public sector services (provision and quality). - Tax morale and social norms. - Deterrence (fines, perceived risk of detection). - Development of official economy (unemployment, GDP growth). - Self-employment rate, size of agricultural sector, cash usage, share of labor force, GDP per capita/growth. - Empirical references: literature cited throughout (e.g., Thomas (1992), Johnson et al. (1997, 1998), Feld and Schneider (2010), Torgler and Schneider (2009)). ### Direct measurement approaches (micro and national accounts) - Direct approaches listed: - (i) System of National Accounts Statistics – Discrepancy method (NAM); - (ii) Survey technique approach; - (iii) Surveys of company managers; - (iv) Estimation of the consumption-income-gap of households. - SNA classification for non-observed economy (Gyomai and van de Ven): - (i) Underground hidden production: legal activities deliberately concealed. - (ii) Illegal production: production forbidden or unlawful when by unauthorized producers. - (iii) Informal sector production: incorporated enterprises in household sector or very small market units. - (iv) Production of households for own (final) use. - (v) Statistical “underground”: missed productive activities due to statistical system deficiencies. - NAM estimation procedure elements: - Correct reporting biases via imputations. - Use upper-bounded estimates for maximum possible NOE activity. - Conduct special purpose surveys and build small-scale indirect models. - NAM categories mapping: - Economic underground: N1+N6 - Informal (and own account production): N3+N4+N5 - Statistical underground: N7 - Illegal: N2 - Cross-country NAM estimates (16 OECD countries, 2011–2012) — selected values: - Italy: 17.5 percent (of official GDP) - Slovak Republic: 15.6 percent - Poland: 15.4 percent - Norway: 1 percent - Micro surveys — representative household/business examples (2015): - Lithuanian Free Market Institute survey: sample size 6,000 across Belarus, Estonia, Latvia, Lithuania, Poland and Sweden (residents aged 18–75; fieldwork May 22–June 15, 2015). - Undeclared working hours as proportion of normal working hours (2015, weekly): - Sweden: 4.2 percent - Poland: 20.7 percent - Average weekly undeclared hours by respondents with shadow experience: - Poland: 25.5 hours - Lithuania: 16.8 hours - Aggregated shadow wages as proportion of GDP: - Sweden: 1.7 percent - Belarus: 32.8 percent - Poland: 24 percent - Surveys of company managers (Putnins and Sauka (2015); Reilly and Krstic (2017)) — average 2009–2015: - Latvia: 27.8 percent - Estonia: 17.4 percent - Lithuania: 16.4 percent - Observation: decline over 2009–2015 for sample countries. ### Indirect (indicator) macro approaches — methods and limitations - Discrepancy between national expenditure and income statistics: assumes hidden incomes but observable expenditure; sensitive to errors in expenditure measures. - Discrepancy between official and actual labor force: weak indicator due to multiple explanations for participation changes. - Electricity approach (Kaufmann and Kaliberda, 1996): difference between electricity consumption growth and official GDP growth as proxy; limitations include activities not requiring electricity and varying electricity-GDP elasticity. - Transaction approach (Fischer quantity equation): Money*Velocity = Prices*Transactions; requires benchmark shadow share and constant proportionality k; suffers from arbitrary constancy assumptions and payment technology changes. - Currency demand approach (CDA, Cagan/Tanzi): assumes informal transactions are cash-based and raise currency demand; isolates “excess” currency demand controlling for conventional factors and shadow determinants (tax burden, regulation complexity); problems include cashless transactions, shifts between currency and deposits, velocity differences, and base-year assumptions. - MIMIC (Multiple Indicators Multiple Causes): explicit SEM for latent variable using multiple causes and indicators (Loayza, 1996). ### MIMIC model and SEM formalization - MIMIC is a special SEM for latent variables; confirmatory and theory-based. - Formal structure: - Structural model: η = ΓX + ζ - Measurement model: y = Λy η + ε - Notation and components (as in source): - η: latent variable (shadow economy) - X: (q×1) vector of causes - Y: (p×1) vector of indicators - Γ: (1×q) coefficient matrix of causes - Λy: (p×1) coefficient matrix in measurement model - ζ: error term in structural model - ε: (p×1) measurement error vector - Example structural specification (causes vector): - [shadow economy] = [γ1, γ2, γ3, γ4, γ5, γ6, γ7, γ8] × [Share of direct taxation; Share of indirect taxation; Share of social security burden; Burden of state regulation; Quality of state institutions; Tax morale; Unemployment quota; GDP per capita] + [ζ] - Example measurement equations: - Employment Quota = λ1 × Shadow Economy + ε1 - Change of local currency = λ2 × Shadow Economy + ε2 - Average working time = λ3 × Shadow Economy + ε3 - MIMIC estimation procedure summary: 1. Model shadow economy as latent variable; 2. Specify structural relations to causes; 3. Link latent variable to indicators via measurement model; 4. Calibrate/benchmark MIMIC results to obtain absolute figures (currency demand or other methods used for calibration). ### Identification, calibration and benchmarking issues - SEMs produce only relative indices; calibration/benchmarking required for absolute (percent of GDP) figures. - Benchmarking procedure criticized: requires external reliable second approach; sensitivity to choice of benchmark. - Identification problem: - Significant structural coefficients show explanatory variables contribute to variance but do not empirically test existence of computed shadow economy. - Testing policy impacts with MIMIC-derived series built using same causal variables yields trivial significance. - Recommendation (Kirchgaessner): use macro approaches independent from MIMIC causes (e.g., electricity approach) to test causal impacts. ### Hybrid model: currency demand + MIMIC (Dybka et al. 2017) - Core innovation: combine currency demand approach (CDA) and MIMIC via “reverse standardization”. - Reverse standardization: supply MIMIC with panel-structured mean and variance of latent variable from CDA estimates; treat these as given in a restricted full-information maximum likelihood. - Advantages: - Avoids choosing arbitrary external reference point for benchmarking. - Constrains variances to non-negative values, addressing negative variance numerical problems in MIMIC estimation. - Enables scale and unit of measurement and construction of a confidence interval. - ANOVA decomposition: 97.2–98.2 percent of SE variance in the panel is due to the CDA component (between cross-sections); remaining small fraction due to MIMIC fine-tuning. - Methodological steps in Dybka et al. (2017): - Estimate and extend panel CDA using frequent and neglected variables (including electronic payment system development) and abandon assumption that shadow share is zero in base. - Estimate MIMIC by maximizing full-information likelihood reformulated using CDA means/variances and constraining parameters to avoid negative variances. - Empirical comparison summary: - Dybka et al. (2017) estimates are on average much lower than MIMIC macro and MIMIC adjusted ones. - MIMIC adjusted values come close to Dybka et al. (2017) for Bulgaria and Switzerland; country-specific similarities/divergences vary (examples: Bulgaria, Israel, Mongolia, Sweden, UK, Croatia, Moldova). ### The problem of “double counting” in macro approaches - Macro approaches may include: - do-it-yourself activities, - neighbors’ or friends’ help, - legally bought material used in shadow activities, - smuggling. - Example decomposition (average 2009–2015): - Estonia: Macro MIMIC 24.94 percent; after deducting legally bought material, friends’ help, smuggling, DIY and neighbors’ help, corrected shadow economy ≈ 65 percent of macro size → corrected 16.21 percent. - Germany: Macro MIMIC 9.37 percent; corrected ≈ 64.2 percent → corrected 6.02 percent. - Correction factors used: Estonia 65 percent; Germany 64.2 percent. - Adjusted MIMIC sizes are considerably smaller and potentially more realistic for macro methods. ### MIMIC estimation results — scope, patterns, and representative coefficients - MIMIC estimations coverage: 1991–2015 for 158 countries (maximum sample). - Representative model (All Countries, column 1 Table 8) — coefficients and fit: - Trade Openess: -0.086*** - GDP per capita: -0.332*** - Unemployment Rate: 0.051** - Size of Government: 0.102*** - Fiscal Freedom: -0.131*** - Labor Force Participation Rate: -0.521*** - Growth of GDP per capita: -0.208** - Currency (normalization): 1 - RMSEA: 0.073 - Chi-square: 513.407 - Observations: 1897 - Countries: 151 - Significance notation: *** p<0.01, ** p<0.05, * p<0.1. - Consistent empirical patterns: - Trade openness typically negative and often significant. - Higher GDP per capita associated with lower SE. - Unemployment positive. - Larger size of government often positive. - Labor force participation rate negative. ### Robustness checks and methodological alternatives - Night lights intensity (Henderson, Storeygard, and Weil, 2012) used as alternative indicator to GDP per capita/growth to avoid using GDP as both cause and indicator. - MIMIC using light intensity (1991–2015) produces theoretical signs and mostly significant causes and indicators across samples. - Caveat: night lights proxy weak for rural economic activity without additional light. - Predictive Mean Matching (PMM) as missing-data imputation approach: - Assumes missing at random (MAR). - Seven-stage PMM imputation (SAS Proc MI) yields multiple imputed datasets; final estimate is average of five datasets. - PMM advantages: non-parametric estimation step, exploits similarity principle, avoids extrapolating from dissimilar countries, captures imputation uncertainty. - Empirical comparison: Spearman’s rank correlation with MIMIC rankings 61 percent (significant at one percent); when grouped into ranges (“<20 percent,” “20–40 percent,” “>40 percent”) over 60 percent coincidence. - Additional robustness: dropping GDP per capita as cause and growth of GDP per capita as indicator retains core relationships in alternative specifications. ### Comparative empirical findings and cross-method comparisons - Summary statistics (MIMIC, 158 countries, 1991–2015): - Mean: 31.9 - Median: 32.3 - Top reported value in truncated text: Zimbabwe with 60.6 - Cross-country examples — smallest and largest averages: - Three smallest shadow economies reported: Austria 8.9; United States 8.3; Switzerland 7.2. - Examples of large shadow economies: Equatorial Guinea 31.8 percent; Suriname 32.2 percent. - Regional averages (1991–2015): - OECD countries: values below 20 percent. - Sub-Saharan African countries: average values above 36 percent. - Latin American countries: average values above 36 percent. - Trend: significant decline in SE over time across country groups; average decline from 1991 to 2015 was 5.3 percentage points. - MIMIC (macro and adjusted) versus National Accounts Discrepancy Method (16 OECD countries, 2011–2012 average) — selected comparisons and observations: - For most countries MIMIC results larger than Discrepancy method (pronounced examples: Norway, Mexico, Belgium, Israel). - Some MIMIC results close to Discrepancy method (e.g., Austria: NOE 7.5 percent; macro MIMIC 8.4 percent; adjusted MIMIC 5.5 percent). - Adjusted MIMIC narrows differences considerably; large differences remain for some countries (e.g., Norway difference 9.7 percentage points). - Conclusion: criticism that MIMIC estimates are unrealistically large should be reconsidered for adjusted MIMIC results. - MIMIC versus National Accounts in Sub-Saharan Africa (Table 20, eight countries, 2010–2014): - Pattern opposite to OECD: Discrepancy method often considerably higher than MIMIC (including adjusted). - Example: Guinea-Bissau — Discrepancy 53.4 percent; MIMIC 37.6 percent; difference 15.8 percentage points. - MIMIC versus micro survey methods (Baltic countries, 2015): - Estonia 2015: MIMIC adjusted 15.3 percent; survey of firm managers 14.9 percent; pure survey 15 percent. - Latvia 2015: macro MIMIC 16.6 percent; MIMIC adjusted 10.8 percent; survey of firm managers 21.3 percent; pure survey 11.7 percent. - Lithuania 2015: MIMIC adjusted 12.2 percent; Putnins and Sauka 15 percent; pure survey 9.8 percent. - Observation: adjusted MIMIC often close to micro survey estimates; pure macro MIMIC tends to be higher. - Macro versus micro broader comparisons (Czech and Slovak Republics, mostly 2008): - Currency Demand Deposit Ratio (Alm and Embaye (2013)): Czech 23.2 percent; Slovak 25.1 percent. - Consumption-Income-Gap (Lichard et al. (2014)): Czech 17.6 percent; Slovak 22.6 percent. - Deterministic Dynamic Simulation (Elgin and Öztunali (2012)): Czech 16.8 percent; Slovak 16.6 percent. - MIMIC macro (Buehn and Schneider, 2008): Czech 15.2 percent; Slovak 16.0 percent. - Conclusion: methods produce substantially different estimates; micro household survey consumption-income-gap can yield results as high as many macro approaches. ### Methodological insights, robustness and innovations (summary) - Macro approaches provide upper-bound estimates because they include crime, DIY and voluntary activities overlapping with “pure” shadow economy. - MIMIC estimations depend heavily on starting values; using other macro estimates as starting values can propagate problems. - Promising methods: - Dybka et al. (2017) structured hybrid CDA+MIMIC approach (reverse standardization). - Night lights (light intensity) as an indicator alternative to GDP per capita/growth. - Predictive Mean Matching (PMM) imputation to avoid calibration issues. - Robustness: trade openness, unemployment rate, GDP per capita, size of government, fiscal freedom and control of corruption are highly statistically significant in most cases; results robust to using light intensity and to excluding GDP variables. - Open research questions: need for more methodological research, validation procedures, and internationally accepted definition of the shadow economy; theoretical linkages between causes and empirical identification remain open. ### Key table and survey figures (selected exact values) - Undeclared working hours as share of normal hours (2015): - Belarus: 17.1 percent - Estonia: 15.0 percent - Latvia: 18.7 percent - Lithuania: 12.8 percent - Poland: 20.7 percent - Sweden: 4.2 percent - Extent of aggregated shadow wages as proportion of GDP (2015): - Belarus: 32.8 percent - Estonia: 15.0 percent - Latvia: 11.7 percent - Lithuania: 9.8 percent - Poland: 24.0 percent - Sweden: 1.7 percent - Representative MIMIC model fit statistics (All Countries): - RMSEA: 0.073 - Chi-square: 513.407 - Observations: 1897 - Countries: 151 - Cross-country sample examples (average shadow economy, 1991–2015): - United States: Average 8.34; Stand. Dev. 0.82; Median 8.23; Min. 7.00; Max. 9.23 - Germany: Average 11.97; Stand. Dev. 2.07; Median 12.80; Min. 7.75; Max. 14.62 - Sweden: Average 13.28; Stand. Dev. 2.15; Median 12.60; Min. 10.12; Max. 16.66 - India: Average 23.91; Stand. Dev. 3.47; Median 24.84; Min. 17.89; Max. 27.83 - Nigeria: Average 56.67; Stand. Dev. 4.10; Median 56.95; Min. 50.64; Max. 66.61 - Zimbabwe: Average 60.64; Stand. Dev. 4.21; Median 60.58; Min. 52.09; Max. 69.08 ### Policy-relevant implications and recommendations (derived from methodological and empirical findings) - Use multiple methods when possible; no single superior method. - Calibrate SEM/MIMIC outputs using independent macro measures (e.g., CDA, electricity, or hybrid approaches) to obtain absolute estimates and avoid circularity. - Adjust macro estimates to correct for double counting (DIY, friends’ help, legally bought materials, smuggling) for realistic SE size interpretation. - Consider night lights intensity as an alternative indicator to GDP per capita/growth to reduce endogeneity in MIMIC specifications. - Employ hybrid CDA+MIMIC approaches (reverse standardization) and PMM where panel coverage or missing data issues exist. - Prioritize development of internationally accepted definitions and validation procedures to enhance comparability and reduce double counting. - Recognize institutional quality, tax morale, and enforcement perceptions as central policy levers to reduce SE participation. *Source: wp1817.* ### 2. THEORETICAL CONSIDERATIONS ### 2. THEORETICAL CONSIDERATIONS ### Conceptual framework and structural equation - Individuals weigh gains from undiscovered shadow activities against losses if detected; participation in the shadow economy (SE) is: - negatively related to the probability of detection p and potential fines f; - positively related to opportunity costs of remaining formal B, which are positively determined by the burden of taxation T and high labor costs W; - the probability of detection p depends on enforcement actions A and facilitating activities F undertaken by individuals to reduce detection. - The discussion suggests the following structural equation (variables as in the source): - SE depends on p, A, F, f, B, T, W in the form indicated in the source text. ### Definition and scope of the shadow economy - Shadow economic activities (SE) are economic activities and income that circumvent government regulation, taxation or observation. - More narrowly, the shadow economy includes monetary and non-monetary transactions of a legal nature that would generally be taxable if reported. - Activities are deliberately concealed to avoid payment of income, value added or other taxes and social security contributions, or to avoid compliance with legal labor market standards (minimum wages, maximum working hours, safety standards) and administrative procedures. - The shadow economy focuses on productive activities that would normally be included in national accounts but remain underground due to tax or regulatory burdens. - Informal household activities such as do-it-yourself activities and neighborly help are typically excluded from the analysis of the shadow economy. ### Causes and indicators of informality - The literature highlights various causes and indicators determining the size of the shadow economy; these are summarized in a table in the source (not reproduced here). ### Direct measurement approaches (micro and national accounts) - Direct approaches described: (i) Measurement by the System of National Accounts Statistics – Discrepancy method (NAM); (ii) Survey technique approach; (iii) Surveys of company managers; (iv) Estimation of the consumption-income-gap of households. - System of National Accounts classification for non-observed economy (Gyomai and van de Ven): - (i) Underground hidden production: legal activities deliberately concealed from public authorities. - (ii) Illegal production: production forbidden by law or unlawful when carried out by unauthorized producers. - (iii) Informal sector production: incorporated enterprises in the household sector or very small units with market production. - (iv) Production of households for own (final) use. - (v) Statistical “underground”: productive activities missed due to deficiencies in statistical systems. - SNA 2008 definitions cited: - Hidden activities reasons include avoiding payment of income tax, value added or other payments; avoiding social security contributions; avoiding legal standards; avoiding administrative procedures. - Illegal activities include (i) production whose sale/distribution/possession is forbidden, and (ii) activities usually legal but illegal when carried out by unauthorized producers. - National Accounts Method (NAM) estimation procedure elements: - Correct data-source reporting biases via imputations. - Use upper-bounded estimates to assess maximum possible NOE activity for given industrial/product groups. - Conduct special purpose surveys and build small-scale indirect models where direct observation is infeasible. - NAM categories and mapping: - Economic underground: N1+N6 - Informal (and own account production): N3+N4+N5 - Statistical underground: N7 - Illegal: N2 - Cross-country NAM estimates (16 developed OECD countries, 2011–2012) show substantial variation: - Italy: 17.5 percent (of official GDP) - Slovak Republic: 15.6 percent - Poland: 15.4 percent - Norway: 1 percent - Micro approach: Representative surveys - Example surveys (Lithuanian Free Market Institute and partners) conducted between May 22 and June 15, 2015, targeting residents aged 18–75; total sample size 6,000 across Belarus, Estonia, Latvia, Lithuania, Poland and Sweden. - Undeclared working hours as proportion of normal working hours (2015, weekly calculation): - Sweden: 4.2 percent - Poland: 20.7 percent - Average weekly undeclared hours by respondents with shadow experience: - Poland: 25.5 hours - Lithuania: 16.8 hours - Aggregated shadow wages as proportion of GDP: - Sweden: 1.7 percent - Belarus: 32.8 percent - Poland: 24 percent - Micro approach: Surveys of company managers - Putnins and Sauka (2015) and Reilly and Krstic (2017) combine misreported business income and misreported wages as a percentage of GDP. - Provides detailed structure of shadow economy in services and manufacturing. - Results (average 2009 to 2015): - Latvia: 27.8 percent - Estonia: 17.4 percent - Lithuania: 16.4 percent - For all countries in their sample, decline observed over period 2009 to 2015. ### Indirect (indicator) macro approaches - Main indirect approaches include: (i) Discrepancy between national expenditure and income statistics: assumes hidden incomes but observable expenditure; depends on no error in expenditure measures. (ii) Discrepancy between official and actual labor force: treats decline in official participation as increase in shadow economy; weak indicator due to other explanations for participation rate changes. (iii) Electricity approach (Kaufmann and Kaliberda, 1996): uses difference between growth of electricity consumption and growth of official GDP as proxy for shadow economy growth; drawbacks include activities not requiring electricity and varying electricity-GDP elasticity. (iv) Transaction approach (Fischer quantity equation): Money*Velocity = Prices*Transactions, assume Prices*Transactions = k (official GDP + shadow economy); requires benchmark shadow economy share and assumes constant k; weaknesses include arbitrary constancy of k and payment technology changes affecting velocity. (v) Currency demand approach (CDA, Cagan/Tanzi): assumes informal transactions are cash-based and increased shadow economy raises currency demand; isolate “excess” currency demand using conventional factors and shadow determinants (tax burden, regulation complexity); problems include cashless transactions, shifts between currency and deposits, velocity differences, and base-year assumptions. (vi) Multiple Indicators, Multiple Causes (MIMIC) approach: explicitly models several causes and effects of the shadow economy using associations between observable causes and effects of an unobserved latent variable (Loayza, 1996). ### MIMIC model and SEM formalization - MIMIC is a special type of structural equation modeling (SEM) for latent variables; theory-based and confirmatory. - Formal structure (as presented): - Modeling the shadow economy as an unobservable (latent) variable. - Structural model: η = ΓX + ζ (latent variable and its causes) - Measurement model: y = Λy η + ε (link between latent variable and indicators) - Notation: - η: latent variable (shadow economy) - X: (q×1) vector of causes in the structural model - Y: (p×1) vector of indicators in the measurement model - Γ: (1×q) coefficient matrix of the causes in the structural equation - Λy: (p×1) coefficient matrix in the measurement model - ζ: error term in the structural model - ε: (p×1) vector of measurement error in y - Structural equation specification (causes listed in the source): - [shadow economy] = [γ1, γ2, γ3, γ4, γ5, γ6, γ7, γ8] × [Share of direct taxation; Share of indirect taxation; Share of social security burden; Burden of state regulation; Quality of state institutions; Tax morale; Unemployment quota; GDP per capita] + [ζ] - Measurement (indicator) equations example (as in source): - Employment Quota = λ1 × Shadow Economy + ε1 - Change of local currency = λ2 × Shadow Economy + ε2 - Average working time = λ3 × Shadow Economy + ε3 - MIMIC estimation procedure steps summarized: 1. Model shadow economy as an unobservable latent variable; 2. Specify relationships between latent variable and causes in a structural model; 3. Link latent variable to its indicators via a measurement model; 4. Use MIMIC results together with a calibration/benchmarking step to obtain absolute figures (procedure continues in subsequent text). *Source: wp1817 - 2. THEORETICAL CONSIDERATIONS (source PDF provided).* ### 1. The first step is that the shadow economy remains an unobserved phenomenon (latent ### wp1817 - 1. The first step is that the shadow economy remains an unobserved phenomenon (latent ### Estimation approach and conceptual framing - The shadow economy is treated as an unobserved phenomenon (latent variable). - Estimation uses: - causes of illicit behavior, e.g. tax burden and regulation intensity; and - indicators reflecting illicit activities, e.g. currency demand and official work time. - The estimation procedure produces only relative estimates of the size of the shadow economy. *Source: wp1817 - 1. The first step is that the shadow economy remains an unobserved phenomenon (latent; https://www.imf.org/-/media/files/publications/wp/2018/wp1817.pdf)* ### 2. In the second step the currency demand method is used to calibrate the relative estimates ### 2. In the second step the currency demand method is used to calibrate the relative estimates into absolute ones by using absolute values of the currency demand method as starting values for the shadow economy. ### Benchmarking and calibration of SEM/MIMIC estimates - SEMs produce only relative weights and an index that shows dynamics of an unobservable latent variable; calibration/benchmarking is required to obtain “real world” figures. - The benchmarking procedure used to derive “real world” figures of shadow economic activities has been criticized; it requires experimentation and comparison across methods and remains unsettled in the literature. - The literature recognizes limitations but considers SEMs valuable tools for analyzing the shadow economy and views objections as incentives for further research rather than a reason to abandon SEMs. ### Identification problem with MIMIC estimates - MIMIC estimates produce only relative weights; normalization requires a second approach whose reliability is crucial. - Significant test statistics in the structural model show explanatory variables contribute to variance of the constructed variable (shadow economy) but do not empirically test the actual existence of the calculated shadow economy or that used causal variables have a statistically significant impact on the “true” shadow economy. - Using variables that were included in the construction of the shadow economy to test policy impacts (e.g., testing tax reductions using MIMIC-derived shadow economy series built using tax variables) leads to trivial statistical significance. - Recommendation by Kirchgaessner: use macro approaches that measure the shadow economy independently from the causes used in MIMIC (e.g., electricity approach) to test causal impacts (e.g., whether a tax increase leads to a rise in the shadow economy). ### A new structured hybrid model: currency demand + MIMIC (Dybka et al. 2017) - Core innovation: a hybrid procedure that combines the currency demand approach (CDA) and MIMIC via a new identification scheme called “reverse standardization”. - Reverse standardization: supply MIMIC with panel-structured information on the latent variable’s mean and variance from CDA estimates, treated as given in a restricted full-information maximum likelihood. - Avoids choosing an externally estimated reference point for benchmarking or ad hoc identifying assumptions (e.g., unity restriction in measurement equation). - Directly addresses numerical problem of negative variances in MIMIC estimation by constraining variances to non-negative values; non-negativity restriction can materially affect significance, specification decisions and measurement results and may flatten the trajectory of the shadow economy. - ANOVA decomposition result: 97.2–98.2 percent of the SE variance in the panel is due to the CDA component (between cross-sections), with only the small remaining fraction due to MIMIC’s fine-tuning. - Methodological steps in Dybka et al. (2017) hybrid approach: - Estimate and extend a panel version of the CDA-equation using frequent and neglected variables (including development of an electronic payment system) and abandon the assumption that the share of the shadow economy in the total economy is zero. - Estimate a MIMIC model by maximizing a full-information likelihood reformulated by (i) using means and variance estimated in the CDA model instead of anchoring an arbitrary period; and (ii) constraining parameter vector to avoid negative variances of structural and measurement errors. - Implication: provides a scale and unit of measurement, avoids ad hoc corrections, and enables construction of a sensible confidence interval. - Empirical comparison (summary): Dybka et al. (2017) estimates are on average much lower than the MIMIC macro and MIMIC adjusted ones; MIMIC adjusted values come close to Dybka et al. (2017) for Bulgaria and Switzerland; similarities and divergences with statistical offices vary by country (examples: Bulgaria, Israel, Mongolia, Sweden, UK, Croatia, Moldova). ### The problem of “double counting” in macro approaches - Macro approaches (MIMIC, CDA, electricity approach) tend to estimate a “total” shadow economy that includes: - do-it-yourself activities, - neighbors’ or friends’ help, - legally bought material used in shadow activities, - smuggling. - Example decomposition (Estonia and Germany, average 2009–2015): - Macro MIMIC estimates: Estonia 24.94 percent; Germany 9.37 percent. - After deducting legally bought material, friends’ help, smuggling, do-it-yourself activities and neighbors’ help, the corrected shadow economy is about two thirds of macro size: - Estonia correction factor: 65 percent. - Germany correction factor: 64.2 percent. - Use of this correction factor produces an adjusted MIMIC size that is considerably smaller and possibly more realistic for macro methods; results presented for 31 European countries for 2017. ### MIMIC estimation results (overview) - MIMIC estimations cover period 1991–2015 for 158 countries (maximum sample); specific country-sample tables discussed include: - All countries: cause variables (trade openness, unemployment, size of government, fiscal freedom, rule of law, control of corruption, government stability) have theoretically expected signs and most are highly statistically significant; indicator variables also have expected signs and are highly statistically significant. - Developing countries (105): rule of law not statistically significant in specification 1; control of corruption not statistically significant in specification 2; other specifications show significance and expected signs; labor force indicator highly significant. - Advanced countries (26): trade openness not statistically significant in all specifications; most other cause variables significant except government stability and size of government; indicator variables are statistically significant and have expected signs. - Note on regression sample construction: The MIMIC regression includes 151 countries generating coefficients and standard deviations; during calibration eight countries were dropped due to short time series and 15 additional countries were added where driver information permitted estimation, completing list of 158 countries with shadow economy estimates. ### Addressing criticisms: Night Lights Intensity approach - Main criticisms addressed: - (i) use of GDP (GDP per capita and growth of GDP per capita) as cause and indicator variables, - (ii) reliance on another independent study to calibrate standardized values to percent of GDP, - (iii) sensitivity of estimated coefficients to specifications, country sample, and time span. - To address (i), replace GDP per capita and growth of GDP per capita with a night lights approach (Henderson, Storeygard, and Weil, 2012) as an independent proxy for economic activity (light intensity as proxy for “true” economic growth). - MIMIC estimations using light intensity (1991–2015): - All countries: cause variables have theoretical signs and most are highly significant except control of corruption; indicator variables have expected signs and are highly significant. - Developing countries (103): unemployment, rule of law and control of corruption not statistically significant in some specifications; labor force indicator highly significant. - Advanced countries (24): trade openness not statistically significant in all specifications; most other causes statistically significant except government stability; indicator variables statistically significant with expected signs. - Caveat: night lights proxy has weaknesses (e.g., rural economic activity may occur without additional light). ### Alternative procedure: Predictive Mean Matching (PMM) - PMM treats estimation of shadow economy as missing data problem; assumes missing at random (MAR) mechanism. - MAR assumption: probability of missingness can depend on observed covariates but not on the missing shadow economy value itself; validated in practice by presence of survey data for large informal economies (e.g., Niger, Burundi). - PMM matching procedure uses a linear regression to establish similarity; variables include GE, RQ, C, ROL, BF, SE, HDI, E with Y as size of the shadow economy (% of GDP). - Seven-stage PMM imputation (using SAS Proc MI): 1. Random draw from posterior predictive distribution of covariate coefficient matrix 훽 to get 훽*. 2. Predict Y* for all countries using 훽*. 3. Identify observed countries whose predicted Y* are closest to Y* of countries missing data; form matches between Y*iobs and Y*imiss. 4. Assign each missing-data country to a group with similar countries with data. 5. Randomly select a match in each group and assign observed outcome from match to the originally missing country. 6. Repeat steps 1–5 five times, generating five datasets with imputed values. 7. Final estimate is the average of the five datasets. - Advantages: PMM is non-parametric in the estimation step, exploits similarity principle to avoid extrapolating from dissimilar countries, and produces multiple imputed datasets capturing uncertainty. - Empirical comparison: PMM results are consistent with MIMIC rankings with Spearman’s rank correlation at 61 percent (significant at one percent); when grouped into “<20 percent of GDP,” “20–40 percent of GDP,” and “>40 percent of GDP,” over 60 percent of countries coincide between samples. ### Additional robustness checks - Robustness test dropping both GDP per capita as cause and growth of GDP per capita as indicator from regressions: MIMIC estimation results for 1991–2015 (tables 15–17) with six alternative specifications per table are consistent with previous results. ### Summary results on size of the shadow economy (158 countries, MIMIC) - Mean value of the size of the shadow economy across 158 countries: 31.9 - Median value: 32.3 - Top reported value from the truncated text: Zimbabwe with 60.6 *Source: IMF Working Paper (content unit: "2. In the second step the currency demand method is used to calibrate the relative estimates into absolute ones by using absolute values of the currency demand method as starting values for the shadow economy").* ### 62.3 and Georgia with 64.9. The three smallest shadow economies are Austria with 8.9, the ### wp1817 - 62.3 and Georgia with 64.9. The three smallest shadow economies are Austria with 8.9, the ### Regional and cross-country patterns in shadow economies - Three smallest shadow economies: Austria with 8.9, the United States with 8.3 and Switzerland with 7.2. - Examples of large shadow economies: Equatorial Guinea with 31.8 percent and Suriname with 32.2 percent of official GDP. - Regional averages (1991–2015): - OECD countries: values below of 20 percent (by far the lowest). - Sub-Saharan African countries: average values above 36 percent. - Latin American countries: average values above 36 percent. - Trend: In all country groups a significant decline in the size of the shadow economy over time; the average decline from 1991 to 2015 was 5.3 percentage points. - By income group: High income countries have the lowest shadow economy and low income countries have the highest. ### Disaggregated results caveats and data notes - Refugee inflows: Many countries, specifically those in the middle east, have been affected by massive refugee inflows; the model does not capture this dimension and therefore the shadow economy in countries such as Jordan, Lebanon and Turkey could potentially be underestimated. For the same reason Syria’s last five year results should be taken with caution. - China: Results should be taken with caution, as it is partly a market economy and partly a planned economy. Therefore, the results might be capturing the informal economy only partially. - Detailed country-year results: For a detailed presentation of the results over all countries and all years see Table A.1 of the Appendix. ### MIMIC (macro and adjusted) versus National Accounts – Discrepancy Method - Table 4.1 (16 OECD countries, years 2011 and 2012 averages) findings: - For most countries, the MIMIC results are considerably larger than the National Accounts Discrepancy method; pronounced examples include Norway, Mexico, Belgium and Israel. - Some MIMIC estimates are close to the Discrepancy method: - Austria: National Accounts Discrepancy method 7.5 percent; macro MIMIC 8.4 percent; adjusted MIMIC 5.5 percent. - Czech Republic and Slovak Republic: somewhat close but macro MIMIC results are considerably higher. - Comparing MIMIC adjusted and Discrepancy method narrows differences considerably. - Specific large differences noted: Norway with 9.7 percentage points (difference), Slovak Rep. with –7.9 percentage points (MIMIC adjusted lower than National Accounts), Belgium with 7.1 percentage points. - Interpretation: - Considerable differences remain between macro MIMIC and Discrepancy methods. - Variance in Discrepancy method is quite large; for at least two or three countries MIMIC comes close. - Prior strong claims that Schneider’s estimates are on average three times as large as System of National Accounts estimates and 6.7 times larger than underground economy estimates should be reconsidered. - Criticism that macroeconomic MIMIC models produce unrealistically large sizes, due to unrealistic model assumptions and calibration decisions, should be reconsidered at least for the adjusted MIMIC results. ### MIMIC versus National Accounts in Sub-Saharan Africa - Table 20 (eight Sub-Saharan African countries, 2010–2014) findings: - Opposite pattern to OECD comparison: for most countries the Discrepancy method is considerably higher than the MIMIC results (including MIMIC adjusted). - In seven out of eight countries the MIMIC estimation is considerably lower than the Discrepancy method. - Example: Guinea-Bissau — Discrepancy method 53.4 percent; MIMIC 37.6 percent; difference 15.8 percentage points. - Implication: Criticism that MIMIC estimates are unrealistically large may not hold for these Sub-Saharan African cases. ### MIMIC versus micro survey methods (Baltic countries, 2015) - Comparison methods: survey of firm managers (Putnins and Sauka (2016)), classical survey (Zukauskas and Schneider (2016)), macro MIMIC and adjusted MIMIC. - Estonia (2015): - MIMIC adjusted 15.3 percent of GDP. - Survey of firm managers 14.9 percent. - Pure survey 15 percent. - Latvia (2015): - Macro MIMIC 16.6 percent. - MIMIC adjusted 10.8 percent. - Survey of firm managers 21.3 percent. - Pure survey 11.7 percent. - Lithuania (2015): - MIMIC adjusted 12.2 percent. - Putnins and Sauka 15 percent. - Pure survey 9.8 percent. - General observation: Adjusted MIMIC estimates are often quite close to micro survey estimates; pure macro MIMIC estimates tend to be considerably higher. ### Macro versus micro methods — broader comparisons - Table 21 (Czech and Slovak Republics, mostly 2008) ranking by size of shadow economy: - Currency Demand Deposit Ratio (Alm and Embaye (2013)): 23.2 percent (Czech), 25.1 percent (Slovak). - Consumption-Income-Gap (Lichard et al. (2014)): 17.6 percent (Czech), 22.6 percent (Slovak). - Deterministic Dynamic Simulation (Elgin and Öztunali (2012)): 16.8 percent (Czech), 16.6 percent (Slovak). - MIMIC macro (Buehn and Schneider, 2008): 15.2 percent (Czech), 16.0 percent (Slovak). - Other methods (Statistical Office Discrepancy Method, Currency Deposit Ratio, other Structural MIMIC) are considerably lower than the top four. - Conclusion: Different methods, even within the same broad class (macro or micro), can produce substantially different estimates; micro household survey Consumption-Income-Gap can yield results as high as many macro approaches. ### Methodological insights, robustness and innovations - Macro approaches provide upper bound estimates because they include crime activities, do-it-yourself activities and voluntary activities, which overlap with “pure” shadow economy activities. - MIMIC estimations depend heavily on starting values; using starting values from other macro estimates can perpetuate problems. - Promising new methods: - Structured hybrid approach by Dybka et al. (2017) combining CDA and MIMIC, avoiding several statistical/econometric problems and yielding much lower estimates. - Light intensity approach used as an indicator variable instead of GDP, avoiding the problem that GDP is often used as both cause and indicator. - Predictive Mean Matching (PMM) methodology (Rubin (1987)) employed to avoid calibration problems and produce plausible results. - Robustness tests: - Trade openness, unemployment rate, GDP per capita, size of government, fiscal freedom and control of corruption are highly statistically significant in most cases. - Results robust when using the light intensity approach. - Results robust to dropping GDP and GDP per capita; trade openness, unemployment rate, size of government, fiscal freedom, rule of law and corruption remain statistically significant. - Robustness holds across sub-samples. ### Summary conclusions - MIMIC estimations for 158 countries over 1991 to 2015 produce plausible results comparable to Schneider (2010), Hassan and Schneider (2016) and other studies. - The light intensity indicator is a viable alternative to GDP per capita or GDP growth as an indicator variable. - PMM method avoids calibration problems of prior MIMIC calibrations and produces plausible results. - Overall finding: a declining size and development of the shadow economy from 1991 to 2015, with a continuous decline interrupted in 2008 due to the world economic crisis. ### Open research questions and recommendations - No superior method: all methodologies have advantages and weaknesses; multiple methods should be used when possible. - Need for more research on estimation methodology and country-period specific results. - Development of satisfactory validation procedures for empirical results to better judge plausibility. - Need for an internationally accepted definition of the shadow economy to facilitate comparisons and avoid double counting. - Theoretical linkages: the link between theory and empirical estimation remains unsatisfactory; identification of core causal and indicator variables remains theoretically open. *Source: IMF Working Paper (content extracted from the provided PDF chapter).* ### 6. REFERENCES ### 6. REFERENCES ### Major methodological and measurement works - Abdih, Y. and Medina, L. (2013), Measuring the Informal Economy in the Caucasus and Central Asia, International Monetary Fund, WP/13/137. - Andreoni, J., Erard B. and J. Feinstein (1998), Tax Compliance, Journal of Economic Literature, 36/4, pp. 818–860. - Breusch, T. (2005b), Estimating the underground economy using MIMIC models, Working Paper. Available under: http://econwpa.wustl.edu/eps/em/papers/0507/0507003.pdf. - Breusch, V. (2016), Estimating the Underground Economy using MIMIC models, Journal of Tax Administration, Vol. 2, No. 1. - Cagan, P. (1958), The demand for currency relative to the total money supply, Journal of Political Economy, 66, pp. 302–328. - Feige, E. (1979), How Big is the Irregular Economy?, Challenge, 22(5), pp. 5–13. - Feige, E.L. (1996), Overseas holdings of U.S. currency and the underground economy. In: Pozo, S. (Ed.), Exploring the Underground Economy. W.E. Upjohn Institute for Employment Research, Kalamazoo, MI, pp. 5–62. - Feige, E.L. (2016a), Reflections on the Meaning and Measurement of Unobserved Economies: What do we really know about the “Shadow Economy”?, Journal of Tax Administration, Vol 2:2. - Feige, Edward L. (2016b), Professor Schneider’s Shadow Economy (SSE): What Do We Really Know? A Rejoinder, Journal of Tax Administration, 2/2. - Gyomai, G., van de Ven, P. (2014), The non-observed economy in the system of national accounts, Statistics Brief No. 18. - Little, R. J. A. (1988), Missing-Data Adjustments in Large Surveys, Journal of Business and Economic Statistics, Vol. 6, No. 3, pp. 287–296. - Little, Roderick, JA and Rubin, B. Donald, Statistical Analysis with Missing Data, Wiley Series in Probability and Statistics, Second Edition, 2002. - Rubin, D.B. (1976), Inference and Missing Data, Biometrika, Vol. 63, No. 3, pp.581–592. - Rubin, D.B. (1987), Multiple Imputation for Nonresponse in Surveys, Wiley. - Zellner, A. (1970). Estimation of Regression Relationships Containing Unobservable Independent Variables. International Economic Review, 11(3), 441–454. - Joreskog, K., and A. S. Goldberger (1975), Estimation of a Model with a Multiple Indicators and Multiple Causes of a Single Latent Variable, Journal of American Statistical Association, Vol. 70, 631–639. ### MIMIC, structural equation, and hybrid modeling approaches - Dell’Anno R. (2007), The Shadow Economy in Portugal: An Analysis with the MIMIC Approach, Journal of Applied Economics, 10, pp. 253–277. - Dell’Anno R., Gomez-Antonio, M. and A. Alanon Pardo (2007), Shadow Economy in three different Mediterranean Countries: France, Spain and Greece. A MIMIC Approach, Empirical Economics, 33, pp. 51–84. - Dell’Anno, R. and F. Schneider (2009), A complex approach to estimate shadow economy: the structural equation modelling, in M. Faggnini and T. Looks (eds.), Coping with the Complexity of Economics, Springer, Berlin, pp. 110–30. - Dybka, P., Kowalczuk, M., Olesinski, B., Rozkrut, M. and Torój, A. (2017): Currency demand and MIMIC models: towards a structured hybrid model-based estimation of the shadow economy size, discussion paper presented at the 5th International Conference on the Shadow Economy, Tax Evasion and Informal Labor July 27–July 30 2017, Warsaw School of Economics, Institute of Econometrics. - Breusch, T. (2005a), The Canadian underground economy: An examination of Giles and Tedds, Canadian Tax Journal, 53, pp. 367–391. ### Empirical surveys, country and regional studies - Amendola, A. and R. Dell’Anno (2010), Institutions and Human Development in the Latin America Shadow Economy, Estudios en Derecho y Gobierno, 3/1, pp. 9–25. - Boeschoten, W.C. and M.M.G. Fase (1984), The Volume of Payments and the Informal Economy in the Netherlands 1965–1982. M. Nijhoff, Dordrecht. - Buehn, A., Karmann, A. and Schneider, F. (2009), Shadow Economy and Do-it-Yourself Activities: The German Case, Journal of Institutional and Theoretical Economics (JITE) 165/4, 701–722. - Buehn, A. and Schneider, F. (2013), Size and Development of Tax Evasion in 38 OECD countries: What do we (not) know?, Discussion Paper, Department of Economics, University of Linz, Linz, Austria, June 2013. - Contini, B. (1981), Labor market segmentation and the development of the parallel economy: the Italian experience, Oxford Economic Papers, 33, pp. 401–412. - Del Boca, D. (1981), Parallel economy and allocation of time, Micros (Quarterly Journal of Microeconomics), 4, pp. 13–18. - Del Boca, D. and F. Forte (1982), Recent empirical surveys and theoretical interpretations of the parallel economy in Italy, in: V. Tanzi (ed.), The Underground Economy in the United States and Abroad, Lexington Books, Lexington, MA, pp. 160–178. - Feld, L.P. and C. Larsen (2005), Black Activities in Germany in 2001 and 2004: A Comparison Based on Survey Data, Study no. 12, Copenhagen: Rockwool Foundation Research Unit. - Feld, L.P. and C. Larsen (2009), Undeclared Work in Germany 2001–2007 – Impact of Deterrence, Tax Policy, and Social Norms: An Analysis Based on Survey Data, Springer, Berlin. - Feld, L. and Schneider, F. (2010). Survey on the Shadow Economy and Undeclared Earnings in OECD Countries, German Economic Review, 11(2), 109–149. - Feld, L.P. and Schneider, F. (2016), " Reply to Gebhard Kirchgaessner", German Economic Review, 18/1, pp.112-117. - Mogensen, G.V., Kvist, H.K., Kfrmendi, E. and S. Pedersen (1995), The Shadow Economy in Denmark 1994: Measurement and Results, Study no. 3, The Rockwool Foundation Research Unit, Copenhagen. - Pedersen, S. (2003), The Shadow Economy in Germany, Great Britain and Scandinavia: A Measurement Based on Questionnaire Service, Study No. 10, The Rockwoll Foundation Research Unit, Copenhagen. - Medina, L., A. Jonelis, and M. Cangul, 2017, The Informal Economy in Sub-Saharan Africa: Size and Determinants, International Monetary Fund, WP/17/156. - Vuletin, G.J. (2008), Measuring the Informal Economy in Latin America and the Caribbean, International Monetary Fund, Working Paper No. 08/102. - Reilly, B. and Krstic, G. (2017), " Shadow Economy - is an Enterprise Survey a Preferable Approach?", Forthcoming in Panoeconomicus, 2017. ### Institutional quality, corruption, tax morale, and policy implications - Dreher, A. and F. Schneider (2009), Corruption and the Shadow Economy: An Empirical Analysis, Public Choice, 144/2, pp. 215–277. - Dreher, A., Kotsogiannis, C. and S. McCorriston (2009), How Do Institutions Affect Corruption and the Shadow Economy?, International Tax and Public Finance, 16/4, pp. 773–796. - Friedman, E., Johnson, S., Kaufmann, D. and P. Zoido-Lobatón (2000), Dodging the Grabbing Hand: The Determinants of Unofficial Activity in 69 Countries, Journal of Public Economics, 76/4, pp. 459–493. - Johnson, S., Kaufmann, D. and A. Shleifer (1997), The unofficial economy in transition, Brookings Papers on Economic Activity, Fall, Washington D.C. - Johnson, S., Kaufmann, D. & Zoido-Lobatón, P. (1998a). Regulatory Discretion and the Unofficial Economy, The American Economic Review, 88(2): 387–392. - Johnson, S., Kaufmann D. and P. Zoido-Lobatón (1998b), Corruption, Public Finances and the Unofficial Economy, World Bank Policy Research Working Paper Series No. 2169, The World Bank, Washington, D.C. - Kaufmann, D. and A. Kaliberda (1996), Integrating the unofficial economy into the dynamics of post socialist economies: a framework of analyses and evidence, in: Kaminski, B. (ed.), Economic Transition in Russia and the New States of Eurasia. M.E. Sharpe, London, pp. 81–120. - Torgler, B. and F. Schneider (2009), The Impact of Tax Morale and Institutional Quality on the Shadow Economy, Journal of Economic Psychology, 30/3, pp. 228–245. - Kirchler, E. (2007), The Economic Psychology of Tax Behaviour, Cambridge (UK) University Press, Cambridge. - Feld, L.P. and B.S. Frey (2007), Tax Compliance as the Result of a Psychological Tax Contract: The Role of Incentives and Responsive Regulation, Law and Policy, 29/1, pp. 102–120. ### Theoretical, conceptual, and review contributions - Portes, A. (1996), The informal economy, in: Pozo, S. (ed.), Exploring the Underground Economy, W.E. Upjohn Institute for Employment Research, Kalamazoo, pp. 147–165. - Loayza, N. V. (1996), The economics of the informal sector: a simple model and some empirical evidence from Latin America, Carnegie-Rochester Conference Series on Public Policy, 45, pp. 129–162. - Gerxhani, K. (2003), The informal sector in developed and less-developed countries: A literature survey, Public Choice, 114/3–4, pp. 295–318. - Portes, A. (1996), The informal economy, in: Pozo, S. (ed.), Exploring the Underground Economy, W.E. Upjohn Institute for Employment Research, Kalamazoo, pp. 147–165. - Thomas, J. J. (1992), Informal Economic Activity, LSE, Handbooks in Economics, Harvester Wheatsheaf, London. - Williams, C. C. and Schneider, F. (2016). Measuring the Global Shadow Economy: The Prevalence of Informal Work and Labour. Edward Elgar Publishing, UK. - Schneider, F. and D. Enste (2002), The Shadow Economy: Theoretical Approaches, Empirical Studies, and Political Implications, Cambridge University Press, Cambridge (UK). - Schneider, F. (ed.) (2011), Handbook on the Shadow Economy, Edward Elgar, Cheltenham. - Schneider, F. (2015), Schattenwirtschaft und Schattenarbeitsmarkt: Die Entwicklungen der vergangenen 20 Jahre, Perspektiven der Wirtschaftspolitik, 16/1, pp. 3–25. - Schneider, F. (2016), Comment on Feige’s Paper‚ Reflections on the Meaning and Measurement of Unobserved Economies: What do we really know about the “Shadow Economy?”, Journal of Tax Administration, Vol 2:2, pp. 82–92. - Schneider, Friedrich (2017), Estimating a Shadow Economy: Results, Methods, Problems, and Open Questions, De Gruyter Open, Open Economics 2017/1, pp. 1–29. - Schneider, F., Buehn, A., & Montenegro, C. E. (2010). New Estimates for the Shadow Economies all over the World. International Economic Journal, 24(4), 443–461. - Schneider, F. and C.C. Willams (2013), The Shadow Economy, IEA, London. - Slemrod, J. and Weber, C. (2012), "Evidence of the Invisible: Toward a Credibility Revolution in the Empirical Analysis of Tax Evasion and the Informal Economy", International Tax and Public Finance, 19/1, pp. 25-53. ### Historical, country-specific, and sectoral case studies - Gutmann, P.M. (1977), The subterranean economy, Financial Analysts Journal, 34/1, pp. 24– 27. - MacAfee, K. (1980), A glimpse of the hidden economy in the national accounts, Economic Trends, 136, pp. 81–87. - O’Neill, D.M. (1983), Growth of the underground economy 1950–81: some evidence from the current population survey, Study for the Joint Economic Committee, U.S. Congress Joint Committee Print, U.S. Gov. Printing Office, Washington, D.C., pp. 98–122. - Tanzi, V. (1980), The underground economy in the United States: estimates and implications, Banca Nazionale del Lavoro, 135, pp. 427–453. - Tanzi, V. (1983), The underground economy in the United States: annual estimates, 1930–1980, IMF Staff Papers, 30, pp. 283–305. - Tanzi, V. (1999), Uses and Abuses of Estimates of the Underground Economy, Economic Journal, 109/3, pp. 338–347. - Witte, A.D. (1987), The nature and extent of unreported activity: a survey concentrating on a recent US research, in: Alessandrini, S. and B. Dallago (eds.), The Unofficial Economy: Consequences and Perspectives in Different Economic Systems, Gower, Aldershot. - Yoo, T. and Hyun, J. K. (1998), International Comparison of the Black Economy: Empirical Evidence Using Micro-Level Data, paper presented at 1998 Congress of Int. Institute Public Finance, Cordoba, Argentina. - Henderson, V. J., Storeygard, A. and Weil, D. N. (2012), Measuring Economic Growth from Outer Space, American Economic Review, 102(2): 994–1028. - Losby, J.L., Else, J.F., Kingslow, M.E., Edgcomb, E.L., Malm, E.T. and V. Kao (2002), Informal Economy Literature Review, The Aspen Institute, Microenterprise Fund for Innovation, Effectiveness, Learning and Dissemination, Washington D.C., and ISED Consulting and Research, Newark DE. *6. REFERENCES (wp1817)* ### 7. TABLES AND FIGURES ### 7. TABLES AND FIGURES ### Main causes/indicators determining the shadow economy (Table 1) - Tax and social security contribution burdens - The distortion of the overall tax burden affects labor-leisure choices and may stimulate labor supply in the shadow economy. The bigger the difference between the total labor cost in the official economy and after-tax earnings (from work), the greater the incentive to reduce the tax wedge and work in the shadow economy. This tax wedge depends on social security burden/payments and the overall tax burden, making them key determinants in the existence of the shadow economy. - References: Thomas (1992), Johnson, Kaufmann, and Zoido-Lobatón (1998a,b), Giles (1999a), Tanzi (1999), Schneider (2003, 2005), Dell’Anno (2007), Dell’Anno, Gomez-Antonio and Alanon Pardo (2007) - Quality of institutions or corruption - The quality of public institutions is another key factor in the development of the informal sector. Efficient and discretionary application of the tax code and regulations by the government plays a crucial role in the decision to work off the books, even more important than the actual burden of taxes and regulations. Corrupt bureaucracy tends to be associated with larger unofficial activity; good rule of law increases benefits of being formal. Production in the formal sector benefits from higher provision of productive public services and is negatively affected by taxation, while the shadow economy reacts in the opposite way. Strengthening institutions and aligning fiscal policy with median voter preferences can reduce informality. - References: Johnson et al. (1998a,b), Friedman, Johnson, Kaufmann, and Zoido-Lobatón (2000), Dreher and Schneider (2009), Dreher, Kotsogiannis and McCorriston (2009), Schneider (2010), Teobaldelli (2011), Teobaldelli and Schneider (2012), Amendola and Dell’Anno (2010), Losby et al. (2002), Schneider and Williams (2013), Hassan and Schneider (2016), Williams and Schneider (2016) - Regulations - Regulations (e.g., labor market regulations, trade barriers) reduce freedom of choice in the official economy, increase labor costs in the official economy, and provide incentives to work in the shadow economy. Enforcement, rather than the overall extent of regulation, is the key factor inducing informality. - References: Johnson, Kaufmann, and Shleifer (1997), Johnson, Kaufmann, and Zoido-Lobatón (1998b), Friedman, Johnson, Kaufmann, and Zoido-Lobatón (2000), Kucera and Roncolato (2008), Schneider (2011), Hassan and Schneider (2016) - Public sector services - An increase in the shadow economy may lead to fewer state revenues, reducing the quality and quantity of publicly provided goods and services, which can further incentivize participation in the shadow economy. Countries with higher tax revenues achieved by lower tax rates, fewer laws and regulations, a better rule of law and lower corruption levels should have smaller shadow economies. - References: Johnson, Kaufmann, and Zoido-Lobatón (1998a,b), Feld and Schneider (2010) - Tax morale - Tax compliance is driven by a psychological tax contract: taxpayers are more inclined to pay taxes honestly if they receive valuable public services or if political decisions follow fair procedures. Treatment of taxpayers by tax authorities matters: being treated like partners supports compliance. Better tax morale and stronger social norms reduce the probability of individuals working in the shadow economy. - References: Feld and Frey (2007), Kirchler (2007), Torgler and Schneider (2009), Feld and Larsen (2005, 2009), Feld and Schneider (2010) - Deterrence - Empirical evidence on deterrence effects is limited due to lack of international data on legal backgrounds and audit frequency. Available survey evidence shows fines and punishment do not exert a clear negative influence on the shadow economy, while subjectively perceived risk of detection does. Results are often weak and Granger causality tests can show the size of the shadow economy affecting deterrence. - References: Andreoni, Erard and Feinstein (1998), Pedersen (2003), Feld and Larsen (2005, 2009), Feld and Schneider (2010) - Development of the official economy - The higher (lower) the unemployment quota (GDP growth), the higher the incentive to work in the shadow economy, ceteris paribus. - References: Schneider and Williams (2013), Feld and Schneider (2010) - Self-employment - The higher the rate of self-employment, the more activities can be performed in the shadow economy, ceteris paribus. - References: Schneider and Williams (2013), Feld and Schneider (2010) - Unemployment - The higher the rate of unemployment, the higher the probability to work in the shadow economy, ceteris paribus. - References: Schneider and Williams (2013), Williams and Schneider (2016) - Size of the agricultural sector - The larger the agricultural sector, the more possibilities to work in the shadow economy, ceteris paribus. - Reference: Hassan and Schneider (2016) - Use of cash - The larger the shadow economy, the more cash will be used, ceteris paribus. Mostly measured as M0/M1, or M1/M2, or cash per capita outside the banking sector. - References: Hassan and Schneider (2016), Williams and Schneider (2016) - Share of labor force - The higher the shadow economy, the lower the official labor force participation rate, ceteris paribus. - References: Schneider and Williams (2013), Feld and Schneider (2010) - GDP per capita (economic growth) - A larger shadow economy is associated with more economic activities moving out of the formal economy, hence, it shows a decrease in economic growth, ceteris paribus. - Source note: Additionally, section 4 of the study relies on data on light intensity from outer space as a proxy for the “true” economic growth achieved by countries (Medina, Jonelis, and Cangul (2017) used this approach for Sub-Saharan African countries). ### NOE adjustments by informality type – percent of GDP; 2011–2012 (Table 2) - Austria: Underground N1 + N6 = 2.4 (31.7); Illegal N2 = 0.2 (2.1); Informal sector N3 + N4 + N5 = 1.5 (19.4); Statistical deficiencies N7 = 3.5 (46.8); Total NOE = 7.5 (100) - Belgium: Underground = 3.8 (83.8); Illegal = -; Informal sector = -; Statistical deficiencies = 0.7 (16.2); Total NOE = 4.6 (100) - Canada: Underground = 1.9 (88.2); Illegal = 0.2 (8.2); Informal sector = -; Statistical deficiencies = 0.1 (3.6); Total NOE = 2.2 (100) - Czech Rep.: Underground = 6.3 (77.6); Illegal = 0.4 (4.5); Informal sector = 1.3 (15.6); Statistical deficiencies = 0.2 (2.3); Total NOE = 8.1 (100) - France: Underground = 3.7 (54.7); Illegal = -; Informal sector = 2.9 (42.7); Statistical deficiencies = 0.2 (2.7); Total NOE = 6.7 (100) - Hungary: Underground = 3.1 (27.9); Illegal = 0.8 (7.5); Informal sector = 3.1 (28.6); Statistical deficiencies = 3.9 (36); Total NOE = 10.9 (100) - Israel: Underground = 2.2 (32.6); Illegal = -; Informal sector = 1.4 (21.8); Statistical deficiencies = 3 (45.6); Total NOE = 6.6 (100) - Italy: Underground = 16.2 (92.8); Illegal = -; Informal sector = -; Statistical deficiencies = 1.2 (7.2); Total NOE = 17.5 (100) - Mexico: Underground = 5.5 (34.7); Illegal = -; Informal sector = 10.4 (65.3); Statistical deficiencies = -; Total NOE = 15.9 (100) - Netherlands: Underground = 0.8 (36.6); Illegal = 0.5 (20.1); Informal sector = 0.5 (20); Statistical deficiencies = 0.5 (23.2); Total NOE = 2.3 (100) - Norway: Underground = 0.5 (51.5); Illegal = 0 (0.3); Informal sector = 0.5 (43.8); Statistical deficiencies = 0 (4.4); Total NOE = 1 (100) - Poland: Underground = 12.7 (82.6); Illegal = 0.9 (6); Informal sector = 0 (0); Statistical deficiencies = 1.8 (11.4); Total NOE = 15.4 (100) - Slovak Rep.: Underground = 12.1 (77.3); Illegal = 0.5 (3); Informal sector = 2.9 (18.7); Statistical deficiencies = 0.2 (1); Total NOE = 15.6 (100) - Slovenia: Underground = 3.9 (38.2); Illegal = 0.3 (3.2); Informal sector = 2.8 (27.7); Statistical deficiencies = 3.1 (30.9); Total NOE = 10.2 (100) - Sweden: Underground = 3 (100); Illegal = -; Informal sector = -; Statistical deficiencies = -; Total NOE = 3 (100) - UK: Underground = 1.5 (65.6); Illegal = -; Informal sector = 0.5 (22.9); Statistical deficiencies = 0.3 (11.4); Total NOE = 2.3 (100) - Source: Gyomai and van de Ven (2014, p. 6). ### Undeclared working hours as a proportion of normal working hours; year 2015 (Table 3) - Columns: Friends/relatives in shadow labor market (percent); Average weekly undeclared hours worked by respondents with shadow experience; Average weekly undeclared hours worked for the whole population; Normal average weekly working hours; Undeclared hours as a share of normal hours (percent) - Belarus: Friends/relatives in shadow labor market = 29; Average weekly undeclared hours by respondents with shadow experience = 23.5; Average weekly undeclared hours for the whole population = 6.82; Normal average weekly working hours = 39.8; Undeclared hours as a share of normal hours (percent) = 17.1 - Estonia: Friends/relatives in shadow labor market = 26; Average weekly undeclared hours by respondents with shadow experience = 22.4; Average weekly undeclared hours for the whole population = 5.82; Normal average weekly working hours = 38.9; Undeclared hours as a share of normal hours (percent) = 15.0 - Latvia: Friends/relatives in shadow labor market = 36; Average weekly undeclared hours by respondents with shadow experience = 20.3; Average weekly undeclared hours for the whole population = 7.31; Normal average weekly working hours = 39.1; Undeclared hours as a share of normal hours (percent) = 18.7 - Lithuania: Friends/relatives in shadow labor market = 29; Average weekly undeclared hours by respondents with shadow experience = 16.8; Average weekly undeclared hours for the whole population = 4.87; Normal average weekly working hours = 38.1; Undeclared hours as a share of normal hours (percent) = 12.8 - Poland: Friends/relatives in shadow labor market = 33; Average weekly undeclared hours by respondents with shadow experience = 25.5; Average weekly undeclared hours for the whole population = 8.42; Normal average weekly working hours = 40.7; Undeclared hours as a share of normal hours (percent) = 20.7 - Sweden: Friends/relatives in shadow labor market = 8; Average weekly undeclared hours by respondents with shadow experience = 18.9; Average weekly undeclared hours for the whole population = 1.51; Normal average weekly working hours = 36.3; Undeclared hours as a share of normal hours (percent) = 4.2 - Note: Figures for the experience of friends or relatives in the shadow labor market and average weekly undeclared hours are taken from the survey, while normal average weekly working hours come from the Eurostat Database for the year 2014. In the absence of such data for Belarus, it was estimated as an average of normal working hours for Central and Eastern European countries that belong to the European Union. - Source: Zukauskas and Schneider (2016, p. 128). ### Extent of aggregated shadow wages as a proportion of GDP; year 2015 (Table 4) - Columns: Undeclared hours worked per year (Million hours); Average undeclared hourly wage (Euro); Extent of shadow market (Million Euros); GDP (Million Euros); Extent of shadow employment of GDP Proportion (percent) - Belarus: Undeclared hours worked per year = 2,504; Average undeclared hourly wage = 7.51; Extent of shadow market = 18,816; GDP = 57,300; Extent of shadow employment of GDP Proportion (percent) = 32.8 - Estonia: Undeclared hours worked per year = 289; Average undeclared hourly wage = 10.37; Extent of shadow market = 2,993; GDP = 19,963; Extent of shadow employment of GDP Proportion (percent) = 15.0 - Latvia: Undeclared hours worked per year = 549; Average undeclared hourly wage = 5.03; Extent of shadow market = 2,760; GDP = 23,581; Extent of shadow employment of GDP Proportion (percent) = 11.7 - Lithuania: Undeclared hours worked per year = 540; Average undeclared hourly wage = 6.62; Extent of shadow market = 3,570; GDP = 36,444; Extent of shadow employment of GDP Proportion (percent) = 9.8 - Poland: Undeclared hours worked per year = 11,954; Average undeclared hourly wage = 8.24; Extent of shadow market = 98,554; GDP = 410,845; Extent of shadow employment of GDP Proportion (percent) = 24.0 - Sweden: Undeclared hours worked per year = 541; Average undeclared hourly wage = 13.32; Extent of shadow market = 7,212; GDP = 430,635; Extent of shadow employment of GDP Proportion (percent) = 1.7 - Note: Undeclared hours worked per year are calculated as shadow frequency/100 x average weekly undeclared hours worked by persons who carried out shadow activities x 52 x total population aged 18– - Source: table content. *Source: Schneider (2017).* ### 74. Figures for shadow frequency, average undeclared weekly hours, and average undeclared hourly wage ### 74. Figures for shadow frequency, average undeclared weekly hours, and average undeclared hourly wage ### Data sources and coverage - Shadow frequency, average undeclared weekly hours, and average undeclared hourly wage are taken from the survey. - Population aged 18–74 and GDP at current prices are taken from the Eurostat Database for the year 2014. - Source: Zukauskas and Schneider, 2016. ### Comparative sizes of the shadow economy (selected table highlights) - Baltic countries (Putnins and Sauka with Schneider), 2009–2015 - Estonia: 2009 20.2; 2010 19.4; 2011 18.9; 2012 19.2; 2013 15.7; 2014 13.2; 2015 14.9; Average 2009–2015 17.4 - Latvia: 2009 36.6; 2010 38.1; 2011 30.2; 2012 21.1; 2013 23.8; 2014 23.5; 2015 21.3; Average 2009–2015 27.8 - Lithuania: 2009 17.7; 2010 18.8; 2011 17.1; 2012 18.2; 2013 15.3; 2014 12.5; 2015 15.0; Average 2009–2015 16.4 - Selected comparisons of statistical offices vs. currency demand models (Dybka et al. 2017) — Size of the shadow economy (percent of official GDP) - Bulgaria (2014): Statistical offices 9.90; FGLS 14.40; FGLS44 15.40; FGLS44-AR 9.50; MIMIC-M. 21.60; MIMIC-Adj. 14.04 - Denmark (2012): Statistical offices 1.50; FGLS 7.50; FGLS44 5.60; FGLS44-AR 3.90; MIMIC-M. 15.48; MIMIC-Adj. 10.06 - Moldova (2015): Statistical offices 23.70; FGLS 9.90; FGLS44 11.50; FGLS44-AR 7.30; MIMIC-M. 39.68; MIMIC-Adj. 25.79 - Switzerland (2012): Statistical offices 1.30; FGLS 4.00; FGLS44 4.60; FGLS44-AR 3.40; MIMIC-M. 6.66; MIMIC-Adj. 4.33 - Croatia (2015): Statistical offices 6.90; FGLS 13.30; FGLS44 13.70; FGLS44-AR 8.20; MIMIC-M. 22.96; MIMIC-Adj. 14.92 ### Decomposition of shadow economy activities (Estonia and Germany) - Total shadow economy (MIMIC calibrated by currency demand procedures), average 2009–2015 - Estonia: 24.94 percent of official GDP (100 percent of total shadow economy) - Germany: 9.37 percent of official GDP (100 percent of total shadow economy) - Components (rough estimates) - Legally bought material for shadow economy and DIY activities: Estonia 5.24 (21 percent of total); Germany 1.79 (19.1 percent) - Illegal activities (smuggling etc.): Estonia 1.75 (7 percent); Germany 0.69 (7.4 percent) - Do-it-yourself activities and neighbors’ help (without legally bought material): Estonia 1.75 (7 percent); Germany 0.86 (9.2 percent) - Sum of material + illegal + DIY/help: Estonia 8.73 (35 percent); Germany 3.35 (35.7 percent) - “Corrected” shadow economy (total minus sum above): Estonia 16.21 (65 percent); Germany 6.02 (64.2 percent) - Source: Own calculations based on Enste and Schneider (2006) and Buehn and Schneider (2013), p.12. ### MIMIC model estimation: main determinants and representative coefficients (1991–2015, All Countries) - Representative model (column 1 from Table 8) - Trade Openess: -0.086*** - GDP per capita: -0.332*** - Unemployment Rate: 0.051** - Size of Government: 0.102*** - Fiscal Freedom: -0.131*** - Labor Force Participation Rate: -0.521*** - Growth of GDP per capita: -0.208** - Currency (normalization): 1 - RMSEA: 0.073 - Chi-square: 513.407 - Observations: 1897 - Countries: 151 - Consistent patterns across MIMIC specifications: - Trade openness typically enters negatively and is often statistically significant. - Higher GDP per capita is associated with lower estimated shadow economy. - Unemployment Rate typically enters positively. - Larger Size of Government often enters positively. - Labor Force Participation Rate consistently enters negatively. - Notes: Significance notation in source: *** p<0.01, ** p<0.05, * p<0.1. ### Alternative specifications and robustness checks (high-level) - Separate estimations for Developing Countries and Advanced Countries show broadly similar signs though magnitudes and significance vary. - Replacing GDP with night lights (Lights (GDP)) in indicators produces negative coefficients for Lights (GDP) in many specifications (e.g., Lights (GDP) -0.346*** in one All Countries specification). - Excluding GDP and GDP per capita retains key relationships: Trade Openess negative, Unemployment positive, Size of Government positive, Fiscal Freedom negative, Labor Force Participation negative (see Tables 15–17). ### Predictive Mean Matching (PMM) ranking of shadow economy size (average over 1991–2015) — sample entries from each range - Less than 20 percent (PMM and MIMIC columns shown verbatim) - Norway: PMM 1) 1 17.1 - Canada: PMM 2.2 13.9 - United Kingdom: PMM 2.3 11.1 - Sweden: PMM 3 16.3 - Netherlands: PMM 2.3 10.8 - Between 20 percent and 30 percent (sample) - Kazakhstan: PMM 20 38.9 - Colombia: PMM 21.3 33.3 - Romania: PMM 26 30.1 - Indonesia: PMM 26.6 24.1 - Bulgaria: PMM 23.3 29.2 - More than 30 percent (sample) - Lebanon: PMM 30 31.6 - Bangladesh: PMM 30.3 33.6 - Côte d'Ivoire: PMM 31.1 43.4 - Tanzania: PMM 33.4 52.2 - Zimbabwe: PMM 44 60.6 - Notes on Table 14: “PMM 1) Average over 1991–2015”; “MIMIC 2) Average over 1991–2015; results from this paper’s MIMIC estimations.” ### Cross-country summary statistics (selected entries from Table 18) - Full sample: 158 countries, period 1991–2015; Table provides Average, Stand. Dev., Median, Min., Max for each country. - Selected country averages and dispersion (exact values) - United States (USA): Average 8.34; Stand. Dev. 0.82; Median 8.23; Min. 7.00; Max. 9.23 - Germany (DEU): Average 11.97; Stand. Dev. 2.07; Median 12.80; Min. 7.75; Max. 14.62 - Sweden (SWE): Average 13.28; Stand. Dev. 2.15; Median 12.60; Min. 10.12; Max. 16.66 - Estonia (EST): Average 23.80; Stand. Dev. 4.23; Median 24.60; Min. 27.52; Max. 30.51 - India (IND): Average 23.91; Stand. Dev. 3.47; Median 24.84; Min. 17.89; Max. 27.83 - Nigeria (NGA): Average 56.67; Stand. Dev. 4.10; Median 56.95; Min. 50.64; Max. 66.61 - Zimbabwe (ZWE): Average 60.64; Stand. Dev. 4.21; Median 60.58; Min. 52.09; Max. 69.08 - China (CHN): Average 14.67; Stand. Dev. 1.88; Median 15.12; Min. 11.74; Max. 16.52 - Brazil (BRA): Average 37.63; Stand. Dev. 2.75; Median 38.47; Min. 32.56; Max. 41.69 - Norway (NOR): Average 14.07; Stand. Dev. 1.73; Median 13.77; Min. 10.47; Max. 16.35 ### Comparisons with other methods and national accounts - OECD countries (16) comparison MIMIC vs. National Accounts Method (NOE), year 2011/2012 (average) - Example differences (MIMIC Adj. minus NOE) - Norway: NOE 1; MIMIC Adj. (3) 10.7; Difference (3)–(1) 9.7 - Mexico: NOE 15.9; MIMIC Adj. (3) 19.4; Difference 3.5 - Italy: NOE 17.5; MIMIC Adj. (3) 16.3; Difference -1.2 - Slovak Rep.: NOE 15.6; MIMIC Adj. (3) 7.7; Difference -7.9 - Sub-Saharan African countries (averages over 2010–2014) — National Accounts vs. MIMIC and MIMIC Adjusted (Differences) - Mali: National Accounts 55; MIMIC 32.3; MIMIC Adjusted 21.0; Differences (2)–(1) -22.7; (3)–(1) -34.0 - Côte d’Ivoire: National Accounts 34; MIMIC 41.9; MIMIC Adjusted 27.2; Differences (2)–(1) 7.9; (3)–(1) -6.8 - Reported correlation between National Accounts and MIMIC: 0.73; Spearman’s Rank Correlation: 0.857*** ### Regional and income-level patterns (figures reported) - Shadow economy by region, averages (percent of GDP) shown across three subperiods (1991–99, 2000–09, 2010–15) — examples extracted from figures: - Sub-Saharan Africa (SSA): 1991–99 42.36; 2000–09 41.12; 2010–15 39.99 (values appear in figure axis labels) - East Asia and Pacific: 1991–99 26.57; 2000–09 27.31; 2010–15 25.79 - OECD: 1991–99 19.44; 2000–09 24.70; 2010–15 24.19 - Shadow economy by income level (figure 3.5): plotted averages by income grouping and period (1991–99; 2000–09; 2010–15) — numeric ticks and plotted group means are present in the source figures. *Source: wp1817 - 74. Figures for shadow frequency, average undeclared weekly hours, and average undeclared hourly wage (pdf) — Own calculations and cited sources within the chapter.* ### 1. APPENDIX ### wp1817 - 1. APPENDIX ### MIMIC model estimation (Figure A.1) - Model fit and sample: - RMSEA: 0.073 - Chi-Square: 513.407 - Observations: 1897 - Countries: 151 - Displayed parameter values (as reported in the figure): - Growth of GDP per capita → (path coefficient displayed): -0.086 - Informal Economy → Currency (M0 M1): 0.051 - Informal Economy → Labor Force Participation: 0.102 - Informal Economy → Unemployment: -0.049 - Informal Economy → (other displayed coefficient): 1 - GDP per capita ← (path coefficient displayed): -0.521 - Size of Government ← (path coefficient displayed): -0.208 - GDP per capita ← (additional coefficient displayed): -0.33 - Source annotation in figure: Own calculations. ### MIMIC model estimation using night lights (Figure A.2) - Figure caption: Shadow Economy Estimation: The MIMIC Model Using Night Lights - Source annotation in figure: Own calculations. ### Cross-country panel of shadow economy size (Table A.1) — overview and sample statistics - Coverage: - Countries: 158 - Period: 1991 to 2015 - Reported average shadow economy across countries: - Part I (1991–2003) — Av. over countries by year: - 1991: 34.51 - 1992: 34.82 - 1993: 35.22 - 1994: 34.89 - 1995: 34.50 - 1996: 34.14 - 1997: 33.81 - 1998: 33.83 - 1999: 33.78 - 2000: 33.26 - 2001: 33.16 - 2002: 33.14 - 2003: 32.73 - Part II (2004–2015) — Av. over countries by year: - 2004: 31.79 - 2005: 31.24 - 2006: 30.41 - 2007: 29.69 - 2008: 28.98 - 2009: 30.56 - 2010: 29.42 - 2011: 28.77 - 2012: 28.33 - 2013: 28.05 - 2014: 27.44 - 2015: 27.78 - Av. over years (2004–2015): 31.77 - Selected country-level examples (shadow economy size in percent; shown exactly as in the table): - Albania: - 1991: 43.18 … 2003: 32.64 - 2004: 31.72 … 2015: 26.21 - Av. over years (2004–2015): 32.72 - China: - 1991: 17.47 … 2003: 15.12 - 2004: 14.31 … 2015: 12.11 - Av. over years (2004–2015): 14.67 - India: - 1991: 28.43 … 2003: 24.84 - 2004: 23.87 … 2015: 17.89 - Av. over years (2004–2015): 23.91 - United States: - 1991: 10.12 … 2003: 8.40 - 2004: 8.43 … 2015: 7.00 - Av. over years (2004–2015): 8.34 - Germany: - 1991: 13.26 … 2003: 13.18 - 2004: 12.80 … 2015: 7.75 - Av. over years (2004–2015): 11.97 - Nigeria: - 1991: 56.95 … 2003: 57.19 - 2004: 56.72 … 2015: 52.49 - Av. over years (2004–2015): 56.67 - Venezuela, RB: - 1991: 32.02 … 2003: 40.03 - 2004: 36.21 … 2015: 33.63 - Av. over years (2004–2015): 33.81 - Zimbabwe: - 1991: 57.35 … 2003: 61.83 - 2004: 63.50 … 2015: 67.00 - Av. over years (2004–2015): 60.64 - Note: Table A.1 provides year-by-year shadow economy estimates for 158 countries for 1991–2003 (Part I) and 2004–2015 (Part II); averages over countries and over years are reported in the table exactly as shown. *Source: Own calculations.* --- _Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1817.pdf_