Measuring the shadow economy

Measuring the shadow economy using company managersTa¯lis J. Putninš a,b,⇑, Arnis Sauka b,⇑a University of Technology, Sydney, AustraliabStockholm School of Economics in Riga, Riga, Latviaa r t i c l e i n f oArticle history:Received 10 October 2013Revised 19 March 2014Available online 18 April 2014JEL classification:E26O17E01P24H26H32Keywords:Shadow economyTax evasionCompany managersNew EU membersGDPa b s t r a c tPutninš, Ta¯lis J., and Sauka, Arnis—Measuring the shadow economy using company managersThis study develops a method that uses surveys of company managers to measure the sizeof a shadow economy. Our method is based on the premise that company managers are themost likely to know how much business income and wages go unreported due to theirunique position in dealing with both of these types of income. We use a range of surveydesign features to maximize the truthfulness of responses. Our method combines estimates of misreported business income, unregistered or hidden employees, and unreportedwages, to arrive at an estimate of the size of a shadow economy as a percentage of GDP.This approach differs from most other studies of shadow economies, which largely focuson using macroindicators. We illustrate the application of our method to three new EUmember countries. We also analyze the factors that influence companies’ participation inthe shadow economy. Journal of Comparative Economics 43 (2) (2015) 471–490. Universityof Technology, Sydney, Australia; Stockholm School of Economics in Riga, Riga, Latvia.2014 Association for Comparative Economic Studies Published by Elsevier Inc. All rightsreserved.1. IntroductionThe size of a shadow economy is an important issue because informal production has a number of negative consequences.First, informal production and tax evasion can create a vicious spiral: individuals go underground to escape taxes and socialwelfare contributions, eroding the tax and social security bases, causing increases in tax rates and/or budget deficits, pushingmore production underground and ultimately weakening the economic and social basis for collective arrangements. Second,tax evasion can hamper economic growth by diverting resources from productive uses (producing useful goods and services)to unproductive ones (mechanisms and schemes to conceal income, monitoring of tax compliance, issuance and collection ofpenalties for non-compliance). Third, informal production can constrain companies’ ability to obtain debt or equity financingfor productive investment because potential creditors/investors cannot verify the true (concealed) cash flows of the company. This can further impede growth. Finally, shadow activities distort official statistics such as GDP, which are importantsignals to policy makers.Like most phenomena that are not directly observable, shadow economies are difficult to measure. Despite decades ofresearch, the literature is yet to arrive at a consensus on what are the best or most reliable methods of measuring a shadowhttp://dx.doi.org/10.1016/j.jce.2014.04.0010147-5967/ 2014 Association for Comparative Economic Studies Published by Elsevier Inc. All rights reserved.⇑ Corresponding authors. Address: UTS Business School, Broadway NSW 2007, P.O. Box 123, Australia.E-mail addresses: talis.putnins@sseriga.edu (T.J. Putninš), arnis.sauka@sseriga.edu (A. Sauka).Journal of Comparative Economics 43 (2015) 471–490Contents lists available at ScienceDirectJournal of Comparative Economicsjournal homepage: www.elsevier.com/locate/jceeconomy. Importantly, the lack of a broadly accepted method of measurement has hampered research efforts aimed atunderstanding shadow economies: their size, their determinants, their relation to the stages of economic development,and their responsiveness to various policy measures. For example, Feige and Urban (2008) examine the main ‘macro’approaches to measuring the shadow economy and find alarmingly varied estimates across the methods and a lack of convergence. This leads them to conclude ‘‘it is time to acknowledge how little we really know about unobserved economiesdespite forty years of effort to measure their size and growth’’ (p. 300) and suggest ‘‘econometricians must be encouragednot only to critique existing macromethods but to develop constructive alternative means of measurement’’ (p. 300). Thispaper aims to do exactly that.This study makes a methodological contribution by developing an index that measures the size of a shadow economy as apercentage of GDP. We do not seek to directly resolve the debate about advantages, disadvantage, accuracy and reliability ofexisting methods; instead, we aim to extend the promising group of ‘direct methods’ by developing and testing a methodthat has several novel features. Our hope is that with continued refinement of existing methods their estimates will beginto converge.We use the term ‘shadow economy’ (also ‘informal’ or ‘unreported’ economy) to refer to all legal production of goods andservices that is deliberately concealed from public authorities.1 In contrast to studies that focus on analyzing aggregatemacroeconomic data, our estimates of the size of a shadow economy are derived from surveys of company managers. Thereasoning for this approach is that those most likely to know how much business income and wages go unreported are thecompany managers that themselves engage in the misreporting and shadow production. In essence, GDP measured using theincome approach is made up of personal income (wages) and corporate income (business profits). Therefore, the shadoweconomy can be measured as the sum of deliberately concealed wages (including wages paid to unregistered workers) andunreported business profits. Our motivation for turning to company managers to measure the shadow economy is they arein the unique position of simultaneously knowing about both of these components of the shadow economy – unreported wagesand unreported business profits. The challenge is eliciting maximally truthful and precise responses. To do this, our methodmakes use of a number of surveying and data collection techniques shown in previous studies to be effective in eliciting moretruthful responses.The shadow economy index combines estimates of misreported business income, unregistered or hidden employees, aswell as unreported ‘‘envelope’’ wages. Our method requires fewer assumptions than most existing methods of measuring theshadow economy, in particular compared to methods based on macroindicators. Our shadow economy index can be usedthrough time or across sectors and economies and thus is a useful tool for evaluating the effectiveness of policy designedto decrease the size of a shadow economy. Our approach has two main advantages compared to methods using aggregatedmacroeconomic data: (i) it produces micro-level (firm-level) estimates, which can be used to examine the structure of theshadow economy and thus guide policies that target problematic sectors or types of firms, or test theories about factors thatinfluence involvement in the shadow economy; and (ii) it is precise about what parts of observed or unobserved productionare included in the estimates. Therefore estimates from our approach may be used in adjusting GDP to account for shadowproduction.A limitation of our approach is that despite the various surveying and data collection techniques that we use to maximizethe truthfulness of responses, some respondents may still provide untruthful responses due to the sensitive nature of thistopic or due to fear of being exposed to their government authorities. As a result, our estimates may understate the true sizeof the shadow economy and can be treated conservatively as lower-bound estimates. Furthermore, our method is moreexpensive to apply than indirect macromethods.Our paper is not the first to use surveys to estimate the size of the shadow economy,2 but it extends this literature in threemain ways. First, to our knowledge, our method is the first to focus specifically on company managers as the source ofinformation about the shadow economy, motivated by the observation that they play a central role in determining both maincomponents of the shadow economy (misreported business income and misreported wages). This allows us to concurrentlyobtain estimates of both components, which when combined produce an estimate of the full shadow economy. This contrastswith the typical survey-based approach of focusing on households or individuals. Second, our method produces an estimate ofthe shadow economy that is both well-defined and expressed as a proportion of true GDP. This is achieved through careful framing of the survey questions to elicit the quantities of interest, and a set of equations that exploit accounting identities within thesystem of national accounts to map survey responses into a proportional measure of the shadow economy. This feature of ourmethod is what gives it potential to be used in adjusting official GDP estimates to account for the shadow economy. Third, ourmethod is transparent, tested, adequately documented (including a questionnaire form), and sufficiently general such that it canbe applied in a large number of countries. Our method only requires the following two sources of non-survey data: (i) a list oflocal companies with contact details (available from databases such as Bureau Van Dijk’s Orbis, or national business registries);and (ii) national accounts estimates of employees’ remuneration and gross operating income of firms (available from nationalstatistics bureaus or agencies such as Eurostat).1 This definition corresponds to what the Organization for Economic Co-operation and Development (OECD) in their comprehensive 2002 handbook‘‘Measuring the Non-observed Economy’’ as well as the System of National Accounts (SNA 1993) refer to as ‘‘underground production’’. It is also consistent withdefinitions employed by other researchers (e.g., the World Bank study of 162 countries by Schneider et al. (2010)).2 For example, see Isachsen and Strom (1985); Mogensen et al. (1995); Kazemier and van Eck (1992); Kim (2003); Hanousek and Palda (2004); Gerxhani(2007).472 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490To illustrate the application of our method and the type of information that it is capable of producing, we use it to measure the size of the shadow economy in three new EU member states: Estonia, Latvia and Lithuania. Our choice of these threecountries is motivated by several considerations. First, anecdotal evidence suggests former Soviet Union countries such asthe Baltic countries have relatively large shadow economies and thus offer a rich setting to study the determinants andeffects of shadow activity. Second, the Baltic countries are often outliers in cross-country shadow economy studies and estimates of their shadow economies are highly varied across different estimation methods. Existing methods fail to agree onthe size of the Baltic shadow economies, suggesting the need for further evidence. Third, the economies of the Baltic countries, both official and unofficial, are rapidly changing and thus allow for interesting analysis of the dynamics of shadoweconomies. Fourth, since 2010, the governments of the Baltic countries have implemented many policy measures designedto combat the shadow economies, allowing analysis of the effectiveness of these policy measures. We survey approximately500 company owners/managers in each of the three countries in 2011, 2012 and 2013, creating a short dynamic panel ofshadow economy estimates between 2009 and 2012.In addition to documenting the size and dynamics of the shadow economies for the three countries, we illustrate the comparatively rich information produced using our method by testing the factors that influence participation in the shadoweconomy. We find that firms that are dissatisfied with the tax system or with the government tend to engage in more shadow activity. This result is consistent with previous research on tax evasion, and has implications for policies to reduce thesize of the shadow economy. We also find that smaller, younger firms engage in proportionally more shadow activity thanlarger, older firms, consistent with the anecdotal evidence that tax evasion is used by firms to gain a competitive edge, andthat having an edge is important in competing in an established market. Finally, the level of tax evasion and deliberate misreporting is responsive to the perceived probabilities of being caught and to the expected penalties for being caught. This lastfinding is consistent with rational expectations models of tax evasion, but until now has received only limited empiricalsupport.The findings of this study also have a number of policy implications. First, tax compliance can be encouraged byaddressing the high level of dissatisfaction with the tax system and with government, e.g., making tax policy more stable and increasing the transparency with which taxes are spent. The level of shadow activity can also be reduced byincreasing the probability of detection, e.g., increasing the number of tax audits, introducing whistle-blower schemesthat provide incentives to report information to authorities about non-compliant companies, and investment in taxevasion detection technology. Second, the relatively large size of shadow economies in the Baltic countries and theirdifferent expansion/contraction trends are likely to cause significant error in official estimates of GDP and its ratesof change.In the next section we review existing methods of measuring the size of the shadow economy. We then describe ourmethod. Section 4 presents results on the size and determinants of the shadow economies in the Baltic countries, andSection 5 contains conclusions.2. Approaches to measuring the shadow economyExisting methods for measuring the shadow economy can be divided into indirect and direct approaches. Indirectapproaches, also sometimes known as ‘macro’ methods, are all built on the notion that economic activity, whetherreported or hidden, leaves behind observable traces or indicators, such as electricity usage, currency usage, transactions,and official labor force participation rates. In essence, these methods use the observable indicators in various econometricspecifications (in some cases also incorporating the causes of shadow activity) to estimate the true level of economic activity. Subtracting recorded economic activity from the estimate of true economic activity gives an estimate of the shadoweconomy.Many of the indirect macromethods rely on a single indicator of economic activity. For example, the transactionsapproach is based on the assumption that the ratio of transactions to GDP is constant.3 The volume of transactions is estimated as the product of the money supply (cash and current deposits) and the velocity of circulation. By identifying a base yearin which the level of shadow economy is likely to be very low (ideally zero), the method tracks the growth in total economicactivity through time by observing growth in the volume of transactions. Subtracting the official GDP gives a measure of theshadow economy. A limitation of the method is that transactions are an imperfect measure of economic activity. For example,some transactions are unrelated to income generation, some money is used as a store of value, cash and deposits are sensitive tointerest and inflation rates, and credit cards and other payment methods are also used for transactions. Similarly, several scholars have questioned the assumption of a constant ratio of transactions to GDP. Critics point out that the way transactions occur,and consequently the volume of transactions per unit GDP, changes through time and varies across countries causing errors inthe shadow economy estimates.The currency demand approach is based on the assumption that shadow transactions are undertaken with cash in orderto avoid leaving evidence that authorities could find.4 Under this assumption, an increase in the size of the shadow economyincreases the demand for cash. The method controls for factors that naturally affect the demand for cash, for example, payment3 For more detailed descriptions of the transactions approach see Feige (1979) and Feige (1996).4 For details and examples of currency demand methods see Cagan (1958), Gutmann (1977), Tanzi (1980), Tanzi (1983), Feige (1989), Johnson et al. (1998),Williams and Windebank (1995), Ahumada et al. (2007) and Feige and Urban (2008).T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 473habits, interest rates, prevalence of payment cards, and so on. This is done by regressing the cash to deposits ratio on the naturaldemand factors together with a set of factors that cause the shadow economy. The ‘excess’ cash demand (the demand that is notexplained by the natural demand factors) is attributed to transactions in the shadow economy. The method is relatively easy toimplement and therefore has been applied to a large number of countries. Criticisms of the method include the fact that not alltransactions in the shadow economy are conducted with cash, many of the factors that contribute the shadow economy such astax morale are often not included due to lack of data, and identifying a base case involving zero shadow activity is difficult andsubjective.The electric consumption method is based on the empirical observation that aggregate electricity usage is closely relatedto the level of economic activity. The method uses aggregate electricity consumption to obtain an estimate of total (officialand unofficial) GDP. As with the previous methods, subtracting official GDP from the estimate of total GDP gives an estimateof the shadow economy.5 Criticisms of the method include the fact that not all shadow economy activities need a considerableamount of electricity, the need to calibrate against a zero shadow economy base case, the tendency for technological improvements or changes in the structure of the economy to affect the amount of electricity used in production, and differences acrosscountries in the elasticity of electricity/GDP.The macromethods described so far rely on a single indicator of economic activity. In contrast, the ‘multiple indicatormultiple cause’ (MIMIC) method simultaneously considers multiple causes of the existence and growth of the shadoweconomy (e.g., tax burden, regulatory burden, attitudes toward paying taxes, unemployment, etc.) and multiple effects orindicators of the shadow economy (e.g., labor force participation rates, weekly hours worked, etc.).6 MIMIC treats the sizeof the shadow economy as a latent (unobserved) variable and specifies structural equations relating the shadow economy toits causes and indicators. Given data on the assumed causes and indicators, the model’s parameters can be estimated by maximum likelihood. From the parameters one can estimate the relative size of a shadow economy in one country or time periodcompared to another. To arrive at an estimate of the actual size of a shadow economy the relative estimates must be calibratedusing an absolute measure of the shadow economy known or assumed to be correct. Often, MIMIC estimates are calibratedusing estimates from the currency demand approach (e.g., Dell’Anno, 2007; Feld and Schneider, 2010; Schneider et al., 2010)making the absolute values sensitive not only to the MIMIC assumptions (and errors) but also the assumptions (and errors)of the currency demand approach.The advantages of the MIMIC approach include its flexibility in that a researcher can include any list of causes and indicators deemed relevant; although this flexibility can also be viewed as a disadvantage because different causes/indicatorslead to different estimates and the choice of which causes/indicators to use is somewhat subjective. Another advantage isthat MIMIC, like other macromethods, is easier and cheaper to apply to a large sample of countries and/or time periods thandirect methods. The MIMIC approach also has a number of disadvantages and has received strong criticism (e.g., Breusch,2005a,b). Its estimates are sensitive to the calibration method and at this stage there is no consensus on the most suitablecalibration method. MIMIC estimates can also be unstable with respect to changes in the sample or model specification (e.g.,Helberger and Knepel, 1988). Finally, like other macromethods, it is difficult to know what components of shadow activityare captured by the method (Buehn and Schneider, 2013).An indirect method used by many statistical bureaus to incorporate some of the shadow economy into official GDP isbased on discrepancies between income and expenditure statistics or between measures of labor supplied by householdsand labor used by companies.7 Because the income measure of GDP should be equal to the expenditure measure of GDP,the difference between the two (before making adjustments for shadow activity) gives some indication of activity in the shadoweconomy. Similarly, the difference between the amount of labor households supply (e.g., from household surveys) and theamount that companies use (e.g., from social security records) gives some indication of unreported employment. A limitationof such approaches is that discrepancies reflect all omissions and errors throughout official statistics not just shadow activity.Furthermore, unless one of the measurements used in calculating the discrepancy captures all economic activity (all reportedplus all unreported) and the other captures only reported activity, the discrepancy is likely to underestimate the size of the shadow economy.In summary, the strength of indirect macromethods is that they can be calculated with relatively little cost and time, andthus relatively easily applied to compare the sizes of shadow economies across many countries or periods of time. Their mainlimitations include: (i) it is not clear which parts of observed or unobserved production these methods include; (ii) thesemethods draw on very simplified and often unrealistic assumptions; and (iii) the methods are not very stable since changesin assumptions can cause significantly different results.The other group, direct methods, draws on direct micro-level observations, such as income audits or surveys. These are themost expensive and time consuming methods, but they manage to overcome many of the limitations that are typical in indirect methods. Direct methods are recommended for situations in which it is important to define exactly what part of production is being estimated, and when stability of the measurement is important. They also have the advantage of providingdetailed information on the structure of the shadow economy across different sectors, regions, enterprises and individuals.5 For details and examples see Kaufmann and Kaliberda (1996), Johnson et al. (1997), Friedman et al. (2000), Lackó (2000), Rosser et al. (2000) and Feige andUrban (2008).6 For further details and examples see Frey and Weck-Hannemann (1984), Giles (1999), Bajada and Schneider (2005), Dell’Anno (2007) and Schneider et al.(2010).7 For details and examples see MacAfee (1980), Petersen (1982), Thomas (1992) and OECD (2002).474 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490Tax audits conducted by the State Revenue Service (Internal Revenue Service in the US) have been used to estimate thelevel of tax evasion, or the ‘tax gap’.8 The tax gap is related to the shadow economy because most activity in the shadow economy involves tax evasion. Tax audits are not perfect in identifying true income, including reported and unreported components.To overcome this limitation Feinstein (1990, 1991) develops ‘detection controlled estimation’ (DCE) methods that estimate theextent of tax evasion accounting for the fact that audits only detect a non-random fraction of all tax evasion. Audits are alsotypically non-random, which can bias estimates unless the estimation methods account for the lack of randomness. Anotherdisadvantage of audits is that they are unlikely to capture unregistered workers.The second group of direct methods is based on surveys. The key assumption in these methods is that the units ofobservation from which responses are collected, for instance, company managers, (i) know and, (ii) are willing to share information on the extent of their involvement in the shadow economy. Kazemier and van Eck (1992) provide an example of amicro-level survey that is used to estimate part of the shadow economy. They survey households about home maintenanceand repair, including questions on the use of unreported labor and evasion of tax on construction materials. Zienkowski(1996) uses survey methods to estimate the amount of unreported labor within an economy. Kim (2003) uses Soviet familybudget survey data containing various household income and expenditure items to estimate the dynamics of the informaleconomy of Soviet households between 1969 and 1990.Survey-based approaches face the risk of underestimating the total size of the shadow economy due to non-response anduntruthful response given the sensitive nature of the topic. This risk can be minimized by employing a number of survey anddata collection techniques shown in previous studies to be effective in eliciting more truthful responses (e.g., Gerxhani,2007; Kazemier and van Eck, 1992; Hanousek and Palda, 2004). These include guarantees of confidentiality, framing the survey as a study of satisfaction with government policy, gradually introducing the most sensitive questions after less sensitivequestions, phrasing misreporting questions indirectly and, in the analysis, controlling for factors that correlate with potentialuntruthful response such as tolerance toward misreporting. Although these techniques help in obtaining more truthfulresponses, some respondents may still provide untruthful responses. For this reason, survey-based estimates of the sizeof a shadow economy are typically regarded as lower-bound estimates (e.g., Feld and Schneider, 2010; Williams,forthcoming). Survey-based methods are also relatively more expensive to apply than indirect macromethods. An advantageof survey-based methods is that they produce micro-level estimates, which characterize the structure of the shadoweconomy and can provide more detailed guidance to policymakers. Furthermore, it is precise about what parts of observedor unobserved production are included in the estimates and therefore may be used in adjusting GDP to account for shadowproduction. Finally, surveys are able to produce evidence on attitudes and perceptions, which can also be useful in guidingpolicy.For more detailed reviews of the different approaches to measuring the shadow economy and discussions of their advantages and disadvantages, see Schneider and Enste (2000), OECD (2002), Slemrod and Webber (2012) and Buehn andSchneider (2013).3. Method3.1. Survey designOur method of measuring the size of the shadow economy draws on surveys of company owners/managers. Thequestionnaire form (see Appendix A) contains four main sections: (i) external influences; (ii) shadow activity; (iii)company and owner characteristics; and (iv) company managers’ attitudes. To increase the response rate andtruthfulness of responses the questionnaire begins with non-sensitive questions about satisfaction with the governmentand tax policy, before moving to more sensitive questions about shadow activity and deliberate misreporting. This‘gradual’ approach is recommended by methodological studies of survey design in the context of tax evasion andthe shadow economy (e.g., Gerxhani, 2007; Kazemier and van Eck, 1992). Further, the survey is framed as a studyof satisfaction with government policy, rather than a study of tax evasion and misreporting (similar to Hanousekand Palda, 2004).In the first survey block, ‘‘external influences’’, respondents are asked to express their satisfaction with the State RevenueService, tax policy, business legislation and government support for entrepreneurs in the respective country. The questionsuse a five point Likert scale, from ‘‘1’’ (‘‘very unsatisfied’’) to ‘‘5’’ (‘‘very satisfied’’). The first section of the questionnaire alsoincludes two questions related to social norms: tolerance toward tax evasion and toward bribery. Previous studies argue thatpeople are likely to engage in more tax evasion when such behavior is tolerated (Baumol, 1990). The measures of toleranceserve a second important role as control variables for possible understating of the extent of shadow activity due to the sensitivity of the topic.The second section of the questionnaire, ‘‘amount of informal business’’, is constructed based on the concepts of productive, unproductive and destructive entrepreneurship by Baumol (1990), assessment of ‘deviance’ or ‘departure from norms’within organizations (e.g., Warren, 2003) and empirical studies of tax evasion in various settings (e.g., Fairlie, 2002; Aidis andVan Praag, 2007). We assess the amount of shadow activity by asking company managers to estimate the degree of8 For example, see Clotefelter (1983), IRS (1983), Feige (1986) and Feinstein (1991).T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 475underreporting of business income (net profits), underreporting of the number of employees, underreporting of salaries paidto employees and the percentage of revenues that firms pay in bribes. The second section of the questionnaire also elicitscompany managers’ perceptions of the probability of being caught for various forms of shadow activity and the severityof penalties if caught deliberately misreporting.We employ the ‘indirect’ approach for questions about informal business, asking company managers about ‘firms in theirindustry’ rather than ‘their firm’.9 This approach is discussed by Gerxhani (2007) as a method of obtaining more truthfulanswers, and is used by Hanousek and Palda (2004), for example. The study conducted by Sauka (2008) shows, that even ifasked indirectly, company owners/managers’ answers can be attributed to the particular respondent or company that therespondent represents.10 Furthermore, experience from Sauka (2008) suggests that phone interviews are an appropriate toolto elicit information about tax evasion.11 Before conducting each interview, the interviewer ensures that they are talking tothe company’s owner (if that person is also the main manager, as is often the case with small companies) or the highest manager (in cases of delegated management).We use the overlapping years (e.g., answers in both the 2013 survey and 2012 survey about the level of shadow activity in2011) to filter out inconsistent responses. This is only possible in instances where a respondent participates in repeatedsurvey rounds. In particular, our filter drops responses when the same respondent in two different survey rounds answersthe same shadow activity questions about the same reference year with a difference of +/20%. This filtering helps increasethe reliability of survey responses used in calculating the Index.The third section of the questionnaire asks company managers about the performance of their companies (percentagechange in net sales profit, sales turnover and employment during the previous year), the education of the companyowner/manager, company age, industry and region. The fourth section of the questionnaire elicits company managers’opinions about why companies evade taxes.3.2. Calculation of the Shadow Economy IndexThe Index measures the size of the shadow economy as a percentage of GDP.12 There are three common methods of measuring GDP: the output, expenditure and income approaches. Our Index is based on the income approach, which calculates GDPas the sum of gross remuneration of employees (gross personal income) and gross operating income of firms (gross corporateincome). Computation of the Index proceeds in three steps: (i) estimate the extent of underreporting of employee remunerationand underreporting of firms’ operating income using the survey responses; (ii) estimate each firm’s shadow production as aweighted average of its underreported employee remuneration and underreported operating income, with the weightsreflecting the proportions of employee remuneration and firms’ operating income in the composition of GDP; and (iii) calculatea production-weighted average of shadow production across firms.In the first step, underreporting of firm i’s operating income, UROperatingIncome i , is estimated directly from the correspondingsurvey question (Q7). Underreporting of employee remuneration, however, consists of two components: (i) underreportingof salaries, or ‘envelope wages’ (Q11); and (ii) unreported employees (Q9). Combining the two components, firm i’s totalunreported proportion of employee remuneration is:13UREmployeeRemunerationi ¼ 1 ð1 URSalaries i Þð1 UREmployees i Þ ð1ÞIn the second step, for each firm we construct a weighted average of underreported personal and underreported corporateincome, producing an estimate of the unreported (shadow) proportion of the firm’s production (income):ShadowProportioni ¼ acUREmployeeRemuneration i þ ð1 acÞUROperatingIncome i ð2Þwhere ac is the ratio of employees’ remuneration (Eurostat item D.1) to the sum of employees’ remuneration and grossoperating income of firms (Eurostat items B.2 g and B.3 g). We calculate ac for each country, c, in each year using data from9 Even when asked indirectly, some company owners/managers choose not to answer sensitive questions about shadow activity. One way to avoid providingtruthful answers to such questions is by simply answering ‘‘0’’ to all of the shadow activity questions, suggesting that no shadow activity of any kind has takenplace during the past two years. We view it as much more likely that these responses reflect avoidance of sensitive questions than truthful opinions andtherefore treat these cases as non-responses, in order to minimize the downward bias in estimates of shadow activity.10 Sauka (2008) used the following approach: in the follow up survey (one year after the initial survey), respondents are ‘reminded’ that in the initial surveythey stated that, for example, the degree of involvement in underreporting business income by ‘their firm’ (not by ‘firms in their industry’ as formulated in theinitial survey) was, for example, 23%. Each respondent is then asked whether the degree of underreporting in their companies is the same this year and if not, towhat extent it has changed. The conclusion from using this method is that respondents tend to state the amount of underreporting in ‘their firm’ when askedabout ‘firms in their industry’.11 Sauka (2008) uses both face-to-face and phone interviews and concludes that willingness to talk about sensitive issues like tax evasion in Latvia does notdiffer significantly between the two methods.12 Two caveats are worth noting: (i) because we do not measure shadow activity in the state (public) sector, our estimates refer to private sector shadowactivity as a percentage of private sector domestic output; and (ii) we do not attempt to measure the ‘‘black economy’’, i.e., the illegal goods and services.13 In deriving the formula we make the simplifying assumption that wages of unreported employees are on average equal to those of reported employees.476 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490Eurostat. Taking a weighted average of the underreporting measures rather than a simple average is important to allow theShadow Economy Index to be interpreted as a proportion of GDP.14In the third step we take a weighted average of underreported production, ShadowProportioni, across a representativesample of firms in country c to arrive at the Shadow Economy Index for that country:INDEXShadowEconomyc ¼ XNci¼1wiShadowProportioni ð3ÞThe weights, wi, are the relative contribution of each firm to the country’s GDP, which we approximate by the relativeamount of wages paid by the firm. Similar to the second step, the weighting in this final average is important to allowthe Shadow Economy Index to reflect a proportion of GDP.15As a final step, we follow the methodology of the World Economic Forum in their Global Competitiveness Report and applya weighted moving average of INDEXShadowEconomy c calculated from the most recent two survey rounds. There are severalreasons for doing this, including: (i) it increases the amount of available information and hence precision of the Indexby providing a larger sample size; and (ii) it makes the results less sensitive to the specific point in time when the surveyis administered. The weighting scheme comprises two overlapping elements: (i) more weight is given to the more recentsurvey round as that contains more recent information (past information is ‘‘discounted’’); and (ii) more weight is plon larger sample sizes as they contain more information.16 Following the approach of the World Economic Forum, for years inwhich there are no previous surveys (the 2009 and 2010 results, which are based on the first survey round conducted in 2011)the Index is simply based on the one survey round. Consequently, the first two annual Index estimates (2009 and 2010) aremore prone to sampling error than subsequent annual estimates, which benefit from larger samples via the moving average.To allow comparisons across countries we apply consistent methodology in calculating the Shadow Economy Index for each ofthe Baltic countries.3.3. Application of method to Baltic countriesWe apply the method described above to Estonia, Latvia and Lithuania. We use stratified random sampling to construct asample that is representative of the population of companies. We obtain information on all active companies in each of thethree Baltic countries from the Orbis database maintained by Bureau Van Dijk. For each country, we form size quintiles(using book value of assets) and take equal sized random samples from each size quintile. In total, a minimum of 500 phoneinterviews are conducted in each of the three Baltic countries in each survey round.In our application of the survey, we assure the interviewee that the interviewer is from an academic institution, and thatwe guarantee complete confidentiality, i.e., that the company and respondent will not be named in any of the analysis orresults. Such and assurance is likely to encourage more truthful responses. While such assurances are credible when the survey is conducted by an academic institution, the promise of confidentiality is less credible if the study is conducted by anagency of the government. In such cases, it may be possible and desirable for the government agency to outsource the studyto a credible academic institution.The surveys are conducted between March and April of each year and contain questions about shadow activity during theprevious two years. For example, the first survey round, conducted in 2011 is used to estimate the size of the shadow economies in 2009 and 2010. The overlap of one year in consecutive survey rounds is used to validate the consistency ofresponses.In the 2011 survey round we conducted a total of 591 phone interviews in Latvia, 536 in Lithuania, and 500 inEstonia. In 2012 survey round we conducted a total of 503 phone interviews in Latvia, 502 in Lithuania, and 500 inEstonia. In the 2013 survey round we conducted a total of 503 phone interviews in Latvia, 501 in Lithuania, and500 in Estonia.4. Results4.1. Size and components of the shadow economy in the Baltic countriesTable 1 reports the size of the shadow economies in Latvia, Lithuania and Estonia as a percentage of GDP in the years2009–2012. The size of the shadow economy in 2009–2011 is considerably higher in Latvia than Estonia and Lithuania14 For example, suppose in an economy wages sum to 80 and corporate income 20, giving true GDP of 100. Suppose that wages are underreported by 50% andcorporate income by 10% giving an official reported GDP of 40 + 18 = 58. In this example the shadow economy is 42% of true GDP, i.e. (100 58)/100. Aweighted average of the two underreporting proportions accurately estimates the size of the shadow economy: (0.8)(50%) + (1 0.8)(10%) = 42%. However,neither of the two underreporting proportions themselves correctly represent the size of the shadow economy (50% and 10%), nor does an equal weightedaverage: (0.5)(50%) + (1 0.5)(10%) = 30%.15 For an example, consider the previous footnote’s example replacing the two sources of income with two firms: a large one that produces income of 80 and asmaller one that produces income of 20.16 For details on this procedure see the Global Competitiveness Report 2011-2012 (Box 3, p. 64), which is available at: http://www3.weforum.org/docs/WEF_GCR_Report_2011-12.pdf.T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 477(e.g., in 2011 the shadow economy is estimated as 30.2%, 18.9%, and 17.1%, in the three countries respectively). Latvia alsostands out from the neighboring countries in that it has experienced the largest reductions in the size of its shadow economyfrom 2009 to 2012, in both absolute and relative terms (from 36.6% in 2009 to 21.1% in 2012). The change from 2011 to 2012is not statistically significant for Lithuania or Estonia (+1.1% and +0.3 accordingly), i.e., the level of shadow economy hasremained approximately unchanged from 2011 to 2012 in both countries. In Latvia, however, the size of shadow economyin 2012 contracted by 9.1% of GDP compared to the level in 2011. This considerable decline in shadow activity follows a similarly notable decrease in the previous year (2011) of 7.9% of GDP. As a result, the estimates in Table 1 suggest that the size ofthe shadow economy in 2012 is no longer considerably higher in Latvia than in Estonia and Lithuania (21.1% compared to19.2% and 18.2%, respectively). The differences between the three countries in 2012 are marginally statistically significant.17Lithuania and Latvia share the similarity that the size of their shadow economies expanded from 2009 to a peak in 2010, followed by a contraction in 2011, whereas in Estonia the shadow economy seems to have followed a more consistent path ofmodest contractions in both 2010 and 2011.The dynamics of the shadow economies in the Baltic countries during the period 2009–2012 appear to correlate withchanges in macroeconomic conditions. For example, in 2009, when the Latvian shadow economy is estimated to be considerably larger than that of Estonia and Lithuania, the macroeconomic conditions in Latvia were also considerably moreadverse: real GDP was falling at a rate of 17.7% p.a. compared to 14.1% in Estonia and 14.8% in Lithuania, and unemployment in Latvia was 17.3% compared to 13.8% in Estonia and 13.7% in Lithuania. In 2012, the shadow economies are estimated as being more similar in size and the macroeconomic conditions also display less disparity: real GDP is growing at arate of 5.6% p.a. in Latvia, 3.2% in Estonia and 3.6% in Lithuania, and unemployment in Latvia is 14.9% compared to 9.8% inEstonia and 13.2% in Lithuania.18 These observations could, in part, reflect a tendency for shadow sector activity to increaseduring difficult times in the business environment; however, our analysis does not provide any evidence of causality betweenmacroeconomic conditions and the level of shadow economy. In fact, due to our relatively short time-series dimension(2009–2012) we do not even formally test for correlation between the size of the shadow economy and macroeconomicconditions.The differences across the three countries are consistent with various country-specific factors. For example, Estonia wasthe first of the three countries to enter the Eurozone on 1 January 2011. Estonia is also the only one of the three countries toshow a decline in the estimated size of its shadow economy in 2010, consistent with need for greater transparency and legality. Similarly, Latvia joined the Eurozone on 1 January 2014, and it is the only country to show a decline in the size of itsshadow economy in the last year of our sample, 2012. Furthermore, Latvia was the only country in our sample to receivea substantial bailout package during the recent crisis from a conglomerate of international organizations (the EuropeanUnion, the International Monetary Fund, and others). The financial assistance was provided in exchange for the Latviangovernment’s commitment to stringent austerity measures and a series of reforms, including deliberate policy measurestargeting the shadow economy. As a result, the Latvian government set up a high-level working group led by the Ministryof Finance State Secretary and comprising members from various ministries, the Treasury, Latvian Federal Police, StateRevenue Service, the Corruption Prevention and Combatting Bureau, as well as experts from academia, the Bank of Latviaand the Central Statistical Bureau. The key outcome from the working group was a package of 66 different policy actionsto combat the informal economy. All but three of proposed policy actions were approved and implemented at differentpoints in time during 2010–2013. Some involved changes to legislation, others were simply programs or actions undertakenby various ministries and/or government organizations/bureaus, working independently or in collaboration. Thus, Latviaexperienced the strongest deliberate policy efforts aimed at reducing shadow sector activity. Consistent with notion thatTable 1Size of the shadow economy in the Baltic countries 2009–2012.Country 2009 2010 2011 2012 2012–2011Estonia 20.2% 19.4% 18.9% 19.2% +0.3%(18.7%, 21.7%) (18.0%, 20.8%) (16.8%, 20.9%) (16.6%, 21.9%) (2.0%, 2.8%)Latvia 36.6% 38.1% 30.2% 21.1% 9.1%(34.3%, 38.9%) (35.9%, 40.3%) (27.6%, 32.7%) (18.5%, 23.6%) (11.7%, 6.5%)Lithuania 17.7% 18.8% 17.1% 18.2% +1.1%(15.8%, 19.7%) (16.9%, 20.6%) (15.2%, 19.0%) (16.4%, 20.1%) (0.7%, 3.0%)This table reports point estimates and 95% confidence intervals for the size of the shadow economies as a proportion of GDP. The last column reports thechange in the relative size of the shadow economy from 2011 to 2012.17 The p-values for independent samples t-tests of the differences in the mean level of shadow economy for the pairs Latvia–Lithuania and Latvia–Estonia are0.08 and 0.33, respectively.18 IMF World Economic Outlook 2010, 2012 and 2013.478 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490the policy measures were effective, Table 1 suggests Latvia experienced the most significant decline in the size of its shadoweconomy from 2010 to 2012. Finally, the significant transshipment activities between Russia and Western Europe may contribute to the shadow economies in the Baltic countries. About 40% of Russian exports to non-Baltic EU member states aretransported through the Baltic countries (Laurila, 2003; Kovács and Spens, 2006). Of the total cargo carried in the Baltic Seaserving the East–West corridor, Estonian ports handle 12%, Latvian ports 28% and Lithuanian ports 9% (Laurila, 2003; Kovácsand Spens, 2006). The higher transshipment volumes in Latvia, followed by Estonia and then Lithuania, are consistent withour estimates suggesting that in all sample years the size of the shadow economy is largest in Latvia, followed by Estonia andthen Lithuania.19The dynamics of the shadow economy estimates, shown in Table 1, are also largely consistent with estimates from otherindirect methods. For example, Schneider (2013) uses an indirect latent variable method (MIMIC) and obtains the samedynamics for Latvia (an increase from 2009 to 2010 and two subsequent years of decreases). Schneider’s estimates also concur with the dynamics of the Index for Estonia and Lithuania, other than the statistically insignificant changes from 2011 to2012.Although the dynamics are similar, the absolute size of the shadow economies estimated in our study differs from thosein Schneider (2013). For example, taking the most recent available year (2012) our estimates suggest the shadow economy inEstonia, Latvia and Lithuania is 19.2%, 21.1% and 18.2%, respectively, whereas Schneider (2013) estimates the correspondingnumbers to be 28.2%, 26.1% and 28.5%. The difference of 5–10% in the estimates from the two different methods could be dueto a number of reasons. First, with indirect methods such as MIMIC it is difficult to define exactly what part of unobservedproduction is estimated. Therefore, the method applied by Schneider (2013) may effectively measure different componentsof unobserved production, despite using the same definition of ‘shadow economy’. A second reason why our estimates arelower than those of Schneider (2013) could be the tendency of survey-based methods to underestimate the size of the shadow economy, and/or the tendency for indirect methods to overestimate the size of the shadow economy. Indeed, otherpapers point out that survey-based methods tend to produce lower estimates than indirect methods (e.g., Feld andSchneider, 2010; Williams, forthcoming; Buehn and Schneider, 2013). However, there is no consensus as to whether the difference is due to survey-based methods underestimating the size of the shadow economy, indirect methods producinginflated estimates, or both (Williams, forthcoming). Feld and Schneider (2010, p.133) suggest that both overestimationand underestimation play a role: ‘‘the estimates from the MIMIC approach can be regarded as the upper bound of the shadoweconomy, and the estimates obtained from the survey approach provide its lower bound.’’Other widely-used methods for estimating the size of a shadow economy include macromethods such as the electric consumption approach and the currency deposit ratio approach. To the best of our knowledge the most recent application ofthese methods to the Baltic countries is Feige and Urban (2008) for the period 1990–2001.20 The period covered by their studydoes not overlap with ours (2009–2012), and therefore comparisons from the two studies are suggestive at best, keeping inmind that the sizes of shadow economies vary through time.21 Using a modified electric consumption approach (MEC)/currencydeposit ratio approach (CDM) Feige and Urban estimate the size of the shadow economies in Estonia, Latvia and Lithuania as14.3%/21.6%, 19.4%/31.6%, and 19.7%/29.0%, respectively. The corresponding estimates from our method for the closest availableyear (2009) are 20.2%, 36.6%, and 17.7%. Our estimates are within the range given by MEC/CDM for Estonia, above the MEC/CDMestimates for Latvia and below the MEC/CDM estimates for Lithuania. Thus, unlike the comparison with MIMIC, estimates fromour method are not consistently higher or lower than those obtained from electric consumption and currency deposit ratiomethods. The differences between our estimates and the MEC/CDM estimates are partly due to the different sample periods,but also partly due to estimation error stemming from misspecification and violated assumptions in the macromodels. Feigeand Urban (2008) find that the MEC/CDM estimates can be highly inconsistent with one another, suggesting their estimationerrors can be large and preclude confidence in the estimates.Finally, we can also compare our estimates with those of other survey-based methods. The European Commission’sSpecial Eurobarometer (No. 284, ‘‘Undeclared work in the European Union’’) is based on approximately 1000 face-to-facesurveys in each of 27 EU member states during 2007.22 A key difference between the Eurobarometer (2007) survey and ourmethod is the way in which the sensitive questions are framed. The Eurobarometer questionnaire asks questions directly aboutindividuals’ undeclared work/income, whereas our questionnaire uses the indirect approach to illicit more truthful answers,asking instead about practices in other companies within the industry. The Eurobarometer survey asks employed individuals‘‘did your employer pay you all or part of your income in the last 12 months in this way (undeclared ‘envelope’ wages)?’’(QB16), and those that answer ‘‘yes’’ are asked what percentage of their income is paid in this form (QB17). The percentageof employed individuals admitting they receive envelope wages is 8% in Estonia, 17% in Latvia and 11% in Lithuania. Multiplyingthe percentages together (QB16 and QB17) gives a rough estimate of the proportion of total gross wages that are not declared:19 The patterns in transhipment volumes and shadow economy levels also mirror the proportions of ethnic Russians in each of the three countries (28.0% inLatvia, 25.6% in Estonia, and 4.9% in Lithuania, as reported by Vadi and Jaakson (2011)). It is possible that ethnic Russians may feel less satsfied with thegovernment, which can in turn influence their tax evasion behavior. In later subsections we investigate differences across the countries with respect toattitudes and satisfaction with the govement.20 For earlier estimates of the shadow economies in the Baltic countries using the electric consumption and currency demand approaches see Schneider andEnste (2000).21 Estimates from Schneider et al. (2010) and Schneider (2013) suggest that the sizes of the Baltic shadow economies in 2010 are all within 4% of their levels inthe year 2000.22 Williams (2008) provides additional analysis of the Eurobarometer survey data for Eastern European countries.T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 4792.5% in Estonia, 7.9% in Latvia and 5.3% in Lithuania. The authors of Eurobarometer (2007) acknowledge that ‘‘the relatively lowvalues measured in Southern Europe and many Eastern European states are surprising’’ (p. 19), pointing out that these are sensitive issues and due to the direct nature of the questions the estimates ‘‘should be interpreted as the lower limit’’ (p. 18). Thequestion about envelope wages in our questionnaire (Q11) is different in that it asks for an estimate of the percentage of truesalaries in the industry that is not declared (paid in the form of envelope wages). Our results for 2009 suggest that on averageenvelope wages are 19.5% of total wages in Estonia, 34.0% in Latvia, and 15.0% in Lithuania. Our estimates of undeclared wagesare much higher than those in the Eurobarometer study. We attribute the bulk of the difference in estimates to the various techniques that we employ to maximize the truthfulness of responses: asking questions in an indirect form, gradually working up tothe sensitive questions rather than starting with them, framing the survey as a study of satisfaction with government ratherthan a survey of shadow economy, assurances of confidentiality, and an academic affiliation with no connection to governmentauthorities.In contrast to indirect methods, our approach is able to provide more detailed information on the components and determinants of the shadow economy, which we turn to now. Fig. 1 illustrates the relative size of the components of the shadoweconomy in each of the three countries. Envelope wages constitute a large proportion of the shadow economies in all threecountries: they account for 52.3% of the shadow economy in Estonia, 42.9% in Latvia and 39.3% in Lithuania. The next largestcomponent is unreported business income, which makes up between 28.5% and 42.7% of the shadow economy depending onthe country. Finally, unreported employees make up the remainder, and account for between 17.6% and 19.2% of the shadoweconomy in each of the countries.4.2. Determinants of shadow activityIn this subsection we examine the factors that influence firms’ decisions to participate in the shadow economy. The literature on tax evasion identifies two main groups of factors that affect the decision to evade taxes and participate in theshadow economy. The first set emerges from rational choice models of the decision to evade taxes (e.g., Allingham andSandmo, 1972; Yitzhaki, 1974).23 In such models individuals or firms weigh up the benefits of evasion in the form of tax savingsagainst the probability of being caught and the penalties that they expect to receive if caught. The higher the probability ofbeing caught or the more severe the consequences, the less attractive it is to evade taxes. Therefore the decision to underreportincome and participate in the shadow economy is affected by the detection rates, the size and type of penalties, firms’ attitudestoward risk-taking and so on.To measure this first set of factors the survey includes questions about the likelihood of being caught for underreportingbusiness profits, number of employees, salaries, as well as involvement in bribery. We also asked company managers to evaluate potential consequences for the firm if it were caught for deliberate misreporting. Slemrod (2007, p. 38) points out thatEstonia: LatviaLithuaniaFig. 1. Components of the shadow economies in each of the Baltic countries, 2012.23 For good surveys of this literature see Andreoni et al. (1998), Sandmo (2005) and Slemrod (2007).480 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490‘‘there has been no compelling empirical evidence addressing how noncompliance is affected by the penalty for detected evasion, as distinct from the probability that a given act of noncompliance will be subject to punishment.’’ The advantage ofmicro-level survey-based approach is that we are able to separate the probability of detection from the penalty for detectedevasion and estimate their individual impacts on the propensity for a firm to evade taxes.Empirical studies find that the actual amount of tax evasion is considerably lower than predicted by rational choice models. For example, Feld and Frey (2002) point out that at with US levels of detection probability and penalties rational choicemodels would predict that everyone should evade taxes. Alternatively, to explain the actual level of tax evasion in the USusing a simple rational choice model, people must have Arrow–Pratt measures of risk aversion of more than 30, whereasexisting field evidence suggests a range of between one and two. The difference between the relatively high levels of tax evasion predicted by rational choice models and the relatively low levels observed in practice is often attributed to the second,broader, set of tax evasion determinants – attitudes and social norms (see Andreoni et al. (1998) and Slemrod (2007) for areview). These factors include perceived justice of the tax system, i.e., attitudes about whether the tax burden andadministration of the tax system are fair, attitudes about how appropriately taxes are spent and how much firms trustthe government. Finally, tax evasion is also influenced by social norms such as ethical values and moral convictions, as wellas fear of feelings of guilt and social stigmatization if caught.Table 2Determinants of firms’ involvement in shadow activity.Model 1 Model 2 Model 3Intercept 30.502*** 24.578*** 22.368***(4.24) (2.85) (5.60)D_EE 2.103(0.68)D_LT 3.026(0.99)Tolerance_TaxEvasion 1.781** 1.932** 2.423***(2.28) (2.38) (3.64)Satisfaction 1.738* 2.022* 1.962**(1.78) (1.91) (2.28)DetectionProbability 2.847*** 3.042*** 2.709***(2.79) (2.81) (2.96)PenaltyForDetection 0.144 0.049(0.15) (0.05)ln(FirmAge) 1.976 0.715(0.98) (0.33)ln(Employees) 1.458** 1.335⁄ 1.265**(2.18) (1.86) (2.21)AverageWage 0.000 0.001(0.27) (0.61)ChangeInProfit 0.014 0.012(1.05) (0.85)D_Wholesale 1.849 1.994 1.608(0.68) (0.71) (0.67)D_Retail 0.256 0.012 1.066(0.09) (0.00) (0.45)D_Services 1.267 1.469 2.321(0.50) (0.54) (1.09)D_Construction 4.482 4.938 3.424(1.39) (1.41) (1.25)D_OtherSector 1.760 3.447 5.035(0.31) (0.57) (1.23)CountryRegion dummy variables No Yes YesAdjusted R-squared 8.9% 12.2% 12.6%This table reports coefficients from regressions of firms’ unreported proportion of production in 2012 (dependent variable) on various determinants ofshadow activity, using the pooled sample of Estonian, Latvian, and Lithuanian firms. D_EE, and D_LT are dummy variables for Estonian and Lithuanian firms,respectively (Latvian firms are the omitted category). Tolerance_TaxEvasion is the firm’s response to Q5, with higher scores indicating more tolerance.Satisfaction is the first principal component of the firm’s responses to Q1–Q4, with higher scores indicating higher satisfaction with the country’s tax systemand government. DetectionProbability and PenaltyForDetection measure the firm’s perception of the probability of being caught for shadow activity and theseverity of penalties conditional on being caught (calculated as the first principal component of responses to Q16(i)–Q16(iv), and the response to Q17,respectively). ln(FirmAge) and ln(Employees) are the natural logarithms of the firm’s age in years and its number of employees. AverageWage is the averagemonthly salary in EUR paid by the firm. ChangeInProfit is the firm’s percentage change in net sales profit from 2011 to 2012. D_Wholesale to D_OtherSectorare sector dummy variables with manufacturing as the omitted category. T-statistics are reported in parentheses.*** Statistical significance at the 1% level.** Statistical significance at the 5% level.* Statistical significance at the 10% level.T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 481We measure two aspects of the second group of tax evasion determinants, namely attitudes toward the governmentand tax system, and social norms/tax morale. We measure firms’ attitudes toward the government and the tax systemusing four questions about their satisfaction with the State Revenue Service, the government’s tax policy, business legislation and the government’s support for entrepreneurs (see Q1–Q4 in Appendix A). We measure social norms usingtwo questions about the extent to which tax evasion and bribery are considered tolerated behaviors (see Q5–Q6 inAppendix A).All of the factors discussed above (detection rates, the size and type of penalties, attitudes and social norms) are likely todiffer across company characteristics such as country, region, sector, size, and age of firm. These company characteristicsmay also influence a company’s propensity to engage in shadow activity and therefore we include these firm characteristicsas explanatory variables when testing for the determinants of shadow activity.We use regression analysis to identify the factors that influence firms’ involvement in the shadow economy. The regression results are reported in Table 2, using data from the most recent survey round (2013). The results from earlier surveyrounds are similar.24 Model 1 includes most of the possible influential factors and dummy variables for Estonian and Lithuanianfirms. Model 2 replaces the country level dummy variables with countryregion dummy variables. Model 3 drops the statistically insignificant determinants (other than tolerance of tax evasion, which is an important control variable).The country dummy variables suggest that the size of the shadow economy is slightly smaller in Estonia and Lithuaniarelative to Latvia after controlling for a range of explanatory factors, although the coefficients are not statistically significant.Tolerance toward tax evasion is positively associated with the firm’s stated level of income/wage underreporting, i.e.,company managers that view tax evasion as a tolerated behavior tend to engage in more informal activity. The measuresof tolerance also serve the important role of controlling for possible understating of the extent of shadow activity (untruthfulresponses) due to the sensitivity of the topic.25The regression coefficients indicate that the effect of perceived detection probabilities and penalties on the tendency forfirms to engage in deliberate misreporting is consistent with the predictions of rational choice models (e.g., Allingham andSandmo, 1972; Yitzhaki, 1974), i.e., the higher the perceived probability of detection and the larger the penalties, thelower the amount of tax evasion and misreporting. The effect of detection probability in particular stands out as being aFig. 2. Company managers’ attitudes regarding tax evasion.24 The regression results using data from earlier survey rounds are not reported here to save space, but are available from the authors upon request.25 For example, consider two firms that underreport income/wages by 40% each, but the first operates in an environment in which tax evasion is consideredhighly unethical and is not tolerated, whereas the second operates in an environment in which tax evasion is relatively tolerated. The first firm might state thatits estimate of underreporting is around 20% (a downward biased response due to the more unethical perception of tax evasion) whereas the second firm mightanswer honestly that underreporting is around 40%. This example illustrates that failure to control for the sensitivity of tax evasion (proxied here by tolerance)can lead to biased comparisons.482 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490particularly strong deterrent of shadow activity. This evidence suggests a possible policy tool for reducing the size of theshadow economies, namely increasing the probability of detection of misreporting. This could be done via an increased number of tax audits, whistle-blower schemes that provide incentives to report information to authorities about non-compliantcompanies, and investment in tax evasion detection technology.The regression results also indicate that a firm’s satisfaction with the tax system and the government is negativelyassociated with the firm’s involvement in the shadow economy, i.e., dissatisfied firms engage in more shadow activity,satisfied firms engage in less. This result is consistent with previous research on tax evasion, and offers an explanationof why the size of the shadow economy is slightly larger in Latvia than in Estonia and Lithuania; namely that Latvianfirms engage in more shadow activity because they are more dissatisfied with the tax system and the government.26Analyzing each of the four measures of satisfaction separately we find that shadow activity is most strongly related todissatisfaction with business legislation and the State Revenue Service, followed by the government’s tax policy and supportfor entrepreneurs.Another strong determinant of involvement in the shadow economy is firm size, with smaller firms engaging in more shadow activity than larger firms. Firm age, although not statistically significant, suggests that younger firms engage in moreshadow activity than older firms. A possible explanation for these two relations is that small, young firms use tax evasionas a means of being competitive against larger and more established competitors. The sector dummy variables are not statistically significant but suggest that firms in the construction sector tend to engage in more shadow activity. There is noevidence of an association between shadow activity and the average wage paid by a firm or a firm’s change in profits (oremployees or turnover).4.3. Company managers’ attitudes regarding shadow activitiesIn surveying company managers we also elicited opinions about various aspects of the shadow economies in theBaltic countries. We believe that these data might be useful to policy makers, at least as complementary information.More specifically, we asked a number of questions about companies’ motivation to participate in tax evasion. Companymanagers were offered various alternatives and asked to assess those on a 1–7 scale, where ‘1’ represents ‘completelyagree’ and ‘7’ represents ‘completely disagree’. The results from the most recent (2013) survey round are summarizedin Fig. 2.Latvian companies are more inclined to emphasize tax evasion as a possible tool to ensure competitiveness (and survival)of the firm. For example, in response to the statement ‘‘to ensure successful performance of a company (including survival) itis much more important to have an appropriate product than to evade taxes’’, both Estonian and Lithuanian companymanagers reported more toward 1 ‘completely agree’ (1.6 and 1.4, respectively), whereas the average response from Latviancompany managers is 3.0. Also, Latvian company managers, relative to Estonian and Lithuanian managers, tend to agree withthe statement that evading taxes is necessary to survive (response scores of 3.9 in Latvia, 5.1 Lithuania and 5.2 in Estonia).Furthermore, Latvian managers are more inclined to link higher levels of tax evasion with lower past performance. Finally,managers from all three Baltic countries seem to agree that performance of their companies very much depends on theeconomic situation in the country.5. ConclusionsWe develop a method of estimating the size of a shadow economy using surveys of company managers – those that aremost likely to know how much business income and wages go unreported. The key features of our method are: (i) employinga range of survey design features shown in previous studies to help illicit truthful answers (e.g., Gerxhani, 2007; Kazemierand van Eck, 1992; Hanousek and Palda, 2004); and (ii) exploiting precise economic relations between components of GDP togo from survey item responses to an estimate of the size of a shadow economy as a percentage of GDP. Our shadow economyindex combines estimates of misreported business income, unregistered or hidden employees, as well as unreported ‘envelope’ wages. Our method can be applied to a large number of countries; its only data requirements are: (i) a list of local companies with contact details (available from databases such as Bureau Van Dijk’s Orbis, or national business registries); and(ii) national accounts estimates of employees’ remuneration and gross operating income of firms (available from nationalstatistics bureaus or agencies such as Eurostat).The main advantages of our method of measuring the shadow economy, compared to macromethods and latent variablemethods, are as follows. First, our method produces micro-level (firm-level) estimates, which can be used to examine howinvolvement in the shadow economy varies across firms. Such information can be very useful to policymakers and regulatorsin targeting problematic sectors or types of firms, or in testing cross-sectional theories about companies’ involvement in theshadow economy. Second, our method is precise about what parts of observed or unobserved production are included in theestimates. As a result, estimates of the shadow economy using our method may be useful in adjusting official GDP measures26 The average satisfaction score in each of the four areas (State Revenue Service, the government’s tax policy, business legislation and the government’ssupport for entrepreneurs) is lower for Latvian company managers than managers in neighboring countries. Dissatisfaction in Latvia is particularly high withthe government’s tax policy.T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 483to account for the shadow economy. This is an important area for future research, given the widespread use of GDP estimatesin measuring the success or failure of economic policies, constructing ratios such as budget deficit to GDP and public sectordebt to GDP, which determine outcomes such as international economic aid and currency union accession, calculating thelevel of contributions and subsidies in economic unions such as the European Union, and in cross-country comparisons ofeconomic development and welfare.A downside of our method is that it is more costly and time-consuming (than macromethods) to collect the necessaryinputs to the calculation. Furthermore, to the extent that some respondents will provide untruthful responses about the shadow economy, despite assurances of confidentiality and various surveying techniques designed to illicit truthful answers, theshadow economy estimates will understate the true size of the shadow economy.We illustrate our method by applying it to the Baltic countries in three successive years, forming a short panel of shadowactivity estimates. Our results indicate some interesting cross-sectional differences and dynamics in the size of the Balticshadow economies. The size of the Latvian shadow economy as a proportion of GDP has seen a considerable contraction from2009 to 2012. In contrast the sizes of the shadow economies in Estonia and Lithuania have remained more stable with modest fluctuations during the past four years. In light of the detrimental long-term effects of having a large shadow economy,the contraction in the Latvian shadow economy is an encouraging sign. Despite its significant decrease, the size of the shadow economy in Latvia remains somewhat larger than in neighboring Lithuania and Estonia, highlighting the importance ofcontinuing policy efforts to minimize shadow sector activities.The dynamics of our shadow economy estimates are consistent with other studies that use different estimation methods,e.g., Schneider (2013). This serves as an indirect validation of the estimates. An advantage of our approach is that we are ableto provide more detailed information on the components and determinants of the shadow economy. For example, the microlevel evidence suggests that the decline in the size of the shadow economy in Latvia has been driven mainly by decreases inunderreporting of business profits, followed by decreasing envelope wages. The retail, services and construction sectors haveseen the most significant improvements. Large firms have also experienced a considerable improvement in the amount ofmisreporting and tax evasion. The increasing satisfaction of Latvian company managers with the State Revenue Service(SRS) is likely to account for some of the reduction in the level of the Latvian shadow economy. A series of policy actionstargeting the shadow economy in Latvia are likely to have contributed to the decline in the size of the Latvian shadoweconomy.Finally, we illustrate one the main comparative advantages of our method of measuring the shadow economy – themicro-level estimates – by testing the cross-sectional determinants of a firm’s propensity to participate in the shadow economy. We find that firms that are dissatisfied with the tax system or the government tend to engage in more shadow activity.This result is consistent with previous research on tax evasion, and has implications for policies to reduce the size of theshadow economy. We also find that smaller, younger firms engage in proportionally more shadow activity than larger, olderfirms, consistent with the anecdotal evidence that tax evasion is used by firms to gain a competitive edge, and that having anedge is important in competing in an established market. Finally, the level of tax evasion and deliberate misreporting amongBaltic companies is responsive to the perceived probabilities of being caught and to the expected penalties for being caught,consistent with rational expectations models of tax evasion. In particular, companies that perceive the probability of beingcaught as being higher tend to engage in less shadow activity.Our findings suggest a number of approaches for policymakers to reduce the size of the shadow economies in countriessimilar to the new EU member states to which we applied our method. First, reducing dissatisfaction with the tax system islikely to decrease the size of the shadow economies. Addressing this issue could involve actions such as making tax policymore stable (less frequent changes in procedures and tax rates), making taxes more ‘fair’ from the perspective of businessesand employees, and increasing the transparency with which taxes are spent. Second, increasing the probability of detectionis expected to reduce shadow activity. This could be achieved via an increased number of tax audits, whistle-blower schemesthat provide incentives to report information to authorities about non-compliant companies, and investment in tax evasiondetection technology.AcknowledgmentsThe authors thank the Centre for Sustainable Business at SSE Riga powered by SEB for the financial support, SKDS for datacollection, the company managers that agreed to participate in interviews, and an anonymous referee for comments thatimproved the paper.484 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490Appendix AT.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 485486 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 487488 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490ReferencesAhumada, Hildegart, Alvaredo, Facundo, Canavese, Alfredo, 2007. The monetary method and the size of the shadow economy: a critical assessment. Reviewof Income and Wealth 53 (2), 363–371.Aidis, Ruta, Van Praag, Mirjam, 2007. Illegal entrepreneurship experience: does it make a difference for business performance and motivation? Analyzingthe effects of illegal entrepreneurship experience in Lithuania. Journal of Business Venturing 22 (2), 283–310.Allingham, Michael G., Sandmo, Agnar, 1972. Income tax evasion: a theoretical analysis. Journal of Public Economics 1, 323–338.Andreoni, James, Erard, Brian, Feinstein, Jonathan, 1998. Tax compliance. Journal of Economic Literature 36, 818–860.Bajada, Christopher, Schneider, Friedrich, 2005. Size, Causes and Consequences of the Underground Economy: An International Perspective. AshgatePublishing Company, Aldershot.Baumol, William, 1990. Entrepreneurship: productive, unproductive and destructive. Journal of Political Economy 98, 893–921.Breusch, Trevor, 2005a. The Canadian underground economy: an examination of Giles and Tedds. Canadian Tax Journal 53 (2), 367–391.Breusch, Trevor, 2005b. Estimating the underground economy using MIMIC models. (Unpublished manuscript).Buehn, Andreas, Schneider, Friedrich, 2013. Estimating the size of the shadow economy: methods, problems and open questions. (Unpublished manuscript).Cagan, Phillip, 1958. The demand for currency relative to the total money supply. Journal of Political Economy 66, 302–328.Clotefelter, Charles T., 1983. Tax evasion and tax rates: an analysis of individual returns. Review of Economic Statistics 65, 363–373.Dell’Anno, Roberto, 2007. The shadow economy in Portugal: an analysis with the MIMIC approach. Journal of Applied Economics 10, 253–277.Eurobarometer, 2007. Special Eurobarometer 284: Undeclared Work in the European Union. European Commission, Brussels.Fairlie, Robert, 2002. Drug dealing and legitimate self- employment. Journal of Labour Economics 20 (3), 538–567.Feige, Edgar L., 1979. How big is the irregular economy? Challenge 22 (5), 5–13.Feige, Edgar L., 1986. A re-examination of the underground economy in the United States. IMF Staff Papers 33 (4), 768–781.Feige, Edgar L., 1989. The Underground Economies: Tax Evasion and Information Distortion. Cambridge University Press, Cambridge, UK.Feige, Edgar L., 1996. Overseas holdings of U.S. currency and the underground economy. In: Pozo, S. (Ed.), Exploring the Underground Economy. W.E. UpjohnInstitute for Employment Research, Kalamazoo, USA, pp. 5–62.Feige, Edgar L., Urban, Ivica, 2008. Measuring underground (unobserved, non-observed, unrecorded) economies in transition countries: can we trust GDP?Journal of Comparative Economics 36, 287–306.Feinstein, Jonathan S., 1990. Detection controlled estimation. Journal of Law and Economics 33, 233–276.Feinstein, Jonathan S., 1991. An econometric analysis of income tax evasion and its detection. Rand Journal of Economics 22, 14–35.Feld, Lars P., Frey, Bruno S., 2002. Trust breeds trust: how taxpayers are treated. Economics of Governance 3, 87–99.Feld, Lars P., Schneider, Friedrich, 2010. Survey on the shadow economy and undeclared earnings in OECD countries. German Economic Review 11 (2), 109–149.Frey, Bruno S., Weck-Hannemann, Hannelore, 1984. The hidden economy as an ‘‘unobserved’’ variable. European Economic Review 26 (1), 33–53.Friedman, Eric, Johnson, Simon, Kaufmann, Daniel, Zoido-Lobaton, Pablo, 2000. Dodging the grabbing hand: the determinants of unofficial activity in 69countries. Journal of Public Economics 76, 459–493.Gerxhani, Klarita, 2007. ‘‘Did you pay your taxes?’’ How (not) to conduct tax evasion surveys in transition countries. Social Indicators Research 80, 555–581.Giles, David, 1999. Modelling the hidden economy in the tax-gap in New Zealand. Empirical Economics 24 (4), 621–640.Gutmann, Peter, 1977. The subterranean economy. Financial Analysts Journal 34, 26–28.Hanousek, Jan, Palda, Filip, 2004. Quality of government services and the civic duty to pay taxes in the Czech and Slovak Republics, and other transitioncountries. Kyklos 57 (2), 237–252.Helberger, Christof, Knepel, Helmut, 1988. How big is the shadow economy? A re-analysis of the unobserved-variable approach of B.S. Frey and H. WeckHannemann. European Economic Review 32 (4), 965–976.IRS. 1983. Income Tax Compliance Research: Estimates for 1973–1981. US Internal Revenue Service Report.Isachsen, Arne J., Strom, Steiner, 1985. The size and growth of the hidden economy in Norway. Review of Income and Wealth 31 (1), 21–38.Johnson, Simon, Kaufmann, Daniel, Schleifer, Andrei, 1997. The unofficial economy in transition. Brookings Papers on Economic Activity 2, 159–239.Johnson, Simon, Kaufmann, Daniel, Zoido-Lobaton, Pablo, 1998. Regulatory discretion and the unofficial economy. American Economic Review 88 (2),387–392.Kaufmann, Daniel, Kaliberda, Aleksander, 1996. Integrating the unofficial economy into the dynamics of post-socialist economies: a framework of analysisand evidence. In: Kaminski, Barlomiej (Eds.), Economic Transition in Russia and the New States of Eurasia. International Politics of Eurasia Series, vol. 8.Sharpe, Armonk and London, pp. 81–120.Kazemier, Brugt, van Eck, Rob, 1992. Survey investigations of the hidden economy. Journal of Economic Psychology 13, 569–587.Kim, Byung-Yeon, 2003. Informal economy activities of Soviet households: size and dynamics. Journal of Comparative Economics 31, 532–551.Kovács, Gyongyi, Spens, Karens M., 2006. Transport infrastructure in the Baltic States post-EU succession. Journal of Transport Geography 14, 426–436.Lackó, Maria, 2000. Hidden economy – An unknown quantity? Comparative analysis of hidden economies in transition economies, 1989–1995. Economicsof Transition 8 (1), 117–149.Laurila, Juhani, 2003. Transit transport between the European Union and Russia in light of Russian geopolitics and economics. Emerging Markets Financeand Trade 39 (5), 27–57.MacAfee, Kerrick, 1980. A glimpse of the hidden economy in the National Accounts. Economic Trends 1980 (8), 81–87.Mogensen, Gunnar V., Kvist, Hans K., Körmendi, Eszter, Pedersen, Soren, 1995. The Shadow Economy in Denmark 1994: Measurement and Results. StudyNo. 3. The Rockwool Foundation Research Unit, Copenhagen.OECD, 2002. Measuring the Non-Observed Economy: A Handbook. OECD, Paris, France.Petersen, Hans-Georg, 1982. Size of the public sector, economic growth and the informal economy: development trends in the Federal Republic of Germany.Review of Income and Wealth 28, 191–215.Rosser, Barkley J., Rosser, Marina V., Ahmed, Ehsan, 2000. Income inequality and the informal economy in transition economies. Journal of ComparativeEconomics 28, 156–171.Sandmo, Agnar, 2005. The theory of tax evasion: a retrospective view. National Tax Journal 58, 643–663.Sauka, Arnis, 2008. Productive, Unproductive and Destructive Entrepreneurship: A Theoretical and Empirical Exploration. Peter Lang GmbH, Frankfurt amMain.Schneider, Friedrich, 2013. Size and development of the shadow economy of 31 European and 5 other OECD countries from 2003 to 2013: a further decline.(Unpublished manuscript).Schneider, Friedrich, Enste, Dominik H., 2000. Shadow economies: size, causes, and consequences. Journal of Economic Literature 38 (1), 77–114.Schneider, Friedrich, Buehn, Andreas, Montenegro, Claudio, 2010. New estimates for the shadow economies all over the world. International EconomicJournal 24 (4), 443–461.Slemrod, Joel, 2007. Cheating ourselves: the economics of tax evasion. Journal of Economic Perspectives 21 (1), 25–48.Slemrod, Joel, Webber, Caroline, 2012. Evidence of the Invisible: Toward a Credibility Revolution in the Empirical Analysis of Tax Evasion in the InformalEconomy. International Tax and Public Finance.Tanzi, Vito, 1980. Underground Economy Built on Illicit Pursuits is Growing Concern of Economic Policymakers. Survey 4-2-1980 International MonetaryFund, pp. 34–37.Tanzi, Vito, 1983. The underground economy in the United States: annual estimates, 1930–1980. IMF Staff Papers 30, 283–305.Thomas, Jeffrey J., 1992. Informal Economic Activity, Handbooks in Economics. Harvester Wheatsheaf, London, UK.T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490 489Vadi, Maaja, Jaakson, Krista, 2011. The dual value of honesty among Russians in selected former Soviet countries. Cross Cultural Management: AnInternational Journal 18 (1), 55–70.Warren, Danielle E., 2003. Constructive and destructive deviance in organizations. Academy of Management Review 28 (4), 622–631.Williams, Colin C., 2008. Illegitimate wage practices in Eastern Europe: the case of ‘‘envelope wages’’. Journal for East European Management Studies 13 (3),253–270.Williams, Colin C., 2013. Out of the shadows: a classification of economies by the size and character of their informal sector. Work, Employment and Society(forthcoming). http://dx.doi.org/10.1177/0950017013501951.Williams, Colin C., Windebank, Jan, 1995. Black market work in the European community: peripheral work for peripheral localities? International Journal ofUrban and Regional Research 19, 23–39.Yitzhaki, Shlomo, 1974. A note on income tax evasion: a theoretical analysis. Journal of Public Economics 3 (2), 201–202.Zienkowski, Leszek, 1996. Polish experience in estimating hidden economy. (Unpublished manuscript).490 T.J. Putninš, A. Sauka / Journal of Comparative Economics 43 (2015) 471–490