Introduction

1Estimating the size of the shadow economy is a difficult task. Any serious attempt to gauge its size should be able to measure a broad range of activities. These activities may include income generated from (say) babysitting, bartering of services, income evaded by (say) a mechanic who chooses to report only part of their income, businesses that overstate their expenses, legitimate income earned and laundered abroad, and income concealed by welfare recipients. Many academic studies have simply focused on some aspects of the vast array of shadow economy activities, and in doing so provide only a lower bound estimate of its size. Part of the problem is the complexity involved in attempting to capture the various facets of these activities. Researchers trying to estimate the shadow economy face the difficulty of how to define it. One commonly used working definition is – the shadow economy comprises of all currently unrecorded economic activities that would otherwise contribute to the officially calculated (or observed) Gross Domestic Product. [1] This conforms to the “production” definition of the shadow economy published in Measuring the Non-Observed Economy: A Handbook – the shadow economy “consist(s) of activities that are productive in an economic sense and quite legal, (provided certain standards and regulations are complied with), but that are deliberately concealed from public authorities”. [2]

2There are many consequences from being substantially involved in shadow economy participation. These include: loss of opportunities and protection that can come from legitimate employment; limited job security; the absence of bargaining rights; limited employment history and thus limited opportunities to secure a good legitimate job in future; the absence of protection with regards to the minimum wage guarantee and other industrial awards; an inability to build up rights to the State pension or contributions to superannuation; difficulties in obtaining a credit card, mortgage, personal loans, etc; difficulty with raising business finance for investment projects when a substantial component of business income is undeclared; the unfair tax burden shouldered by honest tax payers; and the unfair price competition that often results from shadow economic activities. There are also adverse social consequences from the presence of these activities that need careful consideration. Although we do not deal with them explicitly in this paper, it is typical that the economic consequences are given primary attention, leaving by the wayside the social consequences. [3]

3Although there has been considerable discussion on the size of the shadow economy, comparatively little attention has been given to relationship between unemployment and participation in the shadow economy. As Tanzi [1999] points out, “the current literature does not cast much light on these relationships even though the existence of large underground activities would imply that one should look more deeply at what is happening in the labour market” (p. 347). The objective of this paper is to examine the extent of participation in the shadow economy by the unemployed. Knowing something about this will allow us to develop an understanding of how the social security systems in the oecd countries may influence the degree of participation in the shadow economy. To our knowledge this paper provides the first study of its kind. Knowing how the social security system may influence shadow economy participation is also important from a public and social policy perspective. A large and prolonged participation in the shadow economy by the unemployed not only distorts the intended equitable distribution of social security system, it also engenders a possible “dependency trap” if shadow economy income, (when supplemented by social security payments), discourages active participation in the legitime economy. Of course these outcomes depend on the type and duration of financial assistance given to the unemployed, so it is interesting to know the impacts the variations in the social security systems may have.

4The remainder of this paper is organised as follows. In Section 2 we provide the results of estimating the size of the shadow economy, which we use to compare to the rate of unemployment. In Section 3 we examine the various unemployment support programs for a selected number of oecd countries with the objective of determining whether a relationship exists between their generosity and shadow economy participation. In Section 4 we introduce the concepts of an “income effect” and a “substitution effect” to explain the change in employment activities in the shadow economy with the objective of isolating the component of the shadow economy that is influenced directly by changes in the level of unemployment. By examining, in Section 5, the growth cycle characteristics of this component of the shadow economy, we determine how increases and/or decreases in unemployment affect shadow economy activity and the possible effects on unemployment duration that various social transfer programs may have. In section 6, we consider asymmetric (threshold) models to determine whether participation in the shadow economy changes symmetrically or asymmetrically with changes in unemployment. Finally section 7 concludes.

The Shadow Economy and Unemployment in Selected oecd Countries

5The rate of unemployment typically moves counter to contractions and expansions in economic activity. When the economy experiences significant contractions and the unemployment rate increases, this unemployed labour may resurface as clandestine employment in the shadow economy. Although compensated by welfare benefit payments, the decrease in disposable income from being unemployed may foster some dependency on shadow economy activity. However participation in the shadow economy is not confined to just the unemployed. Those who are working in the legitimate economy may also engage in the shadow economy on a part-time basis just as those who are employed legitimately on a part-time basis can afford the time to participate in the shadow economy. It is this heterogeneous group of individuals’ participation in the shadow economy that led Tanzi [1999] to conclude that “relation between the underground economy and the unemployment rate is ambiguous” (p. 343).

6Giles and Tedds [2002] provide a similar explanation for the weak relationship between the two variables. They argue that not only is the unemployment rate negatively correlated with the shadow economy, any increases in unemployment that would coincide with falls in the shadow economy could even contribute positively to shadow economy output “if the unemployed persons spend a portion of their extra time working in the hidden economy” (p. 127). According to Giles and Tedds [2002], the net effect of changes in the unemployment rate on the shadow economy are potentially negligible because of this. This negligible effect does not imply however that it is not important to study the two effects separately. In fact the extent of participation in the shadow economy by the unemployed while in receipt of welfare benefits is important to know as these activities distort the equity intended of the social security system. It is our objective in this paper to provide a method by which to disentangle these two effects on the shadow economy in order to assess: (i) the nature of the relationship between the unemployment rate and shadow economy and (ii) the effects that different social security systems in the oecd countries have on unemployment participation in the shadow economy.

7Our estimates of the shadow economy are derived using the mimic (multiple-indicators multiple-causes) methodology. The mimic approach explicitly considers multiple causes leading to the existence and growth of the shadow economy over time. The empirical method used is quite different from those used elsewhere. It is based on the statistical theory of unobserved variables, which considers multiple causes and multiple indicators of the phenomenon to be measured. For the estimation, a factor-analytic approach is used to measure the shadow economy as an unobserved variable over time. The unknown coefficients are estimated in a set of structural equations within which the “unobserved” variable cannot be measured directly. The mimic model consists in general of two parts, (i) the measurement model that links the unobserved (or latent) variables to observed indicators, and (ii) the structural equations model which specifies causal relationships among the unobserved variables. In this case, there is one unobserved (or latent) variable, the size of the shadow economy. It is assumed to be indirectly observable by a set of indicators of the shadow economy, thus capturing the structural dependence of the shadow economy on variables that may be useful in predicting its movement and size in the future. The reader is referred to Schneider [2006]; Bajada and Schneider [2005]; and Schneider and Enste [2000] for a detailed discussion on the methodology and the causal and indicator variables used in the estimation.

8The results of estimating the shadow economy as a percentage of gdp using the mimic methodology for 12 oecd countries are given in Table 1. [4] In the same table we also report the rate of unemployment for the 12 countries.

Table 1

Estimates of Shadow Economy (% of gdp) and the Unemployment Rate

1991-1997 1998-2005 1991-2005 Shadow Economy (i) Ue rate (ii) Shadow Economy (iii) Ue rate (iv) Shadow Economy (v) Ue rate (vi) Australia 12.58 9.17 13.86 6.33 13.26 7.66 Belgium 20.99 8.63 21.70 7.98 21.37 8.28 Canada 14.34 10.22 15.59 7.39 15.01 8.71 Denmark 15.48 7.41 17.70 4.83 16.66 6.03 Finland 17.22 14.55 17.95 9.52 17.61 11.87 France 12.84 11.33 14.83 9.88 13.90 10.56 Italy 25.14 10.32 26.51 9.41 25.87 9.84 Nz 10.92 8.18 12.34 5.36 11.68 6.67 Norway 17.49 5.21 18.78 3.85 18.18 4.48 Portugal 20.79 5.98 22.48 5.39 21.69 5.66 Sweden 18.49 6.89 18.95 5.12 18.73 5.95 Uk 11.63 8.90 12.51 5.32 12.10 6.99

Estimates of Shadow Economy (% of gdp) and the Unemployment Rate

9The size of the shadow economy and the unemployment rate for the period 1991-2005 are reported in column (v) and (vi) respectively in Table 1. In columns (i) and (iii) we report the average size of the shadow economy for the periods 1991-1997 and 1998-2005 respectively, while in columns (ii) and (iv) we report the average unemployment rate. We observe over this period that the size of the shadow economy increased in each of the 12 countries while the rate of unemployment declined. Although this would suggest that those who are not unemployed are contributing most to shadow economy activities, it does not rule out the possibility that the unemployed are also participating. In addition we also observe from Table 1 that a country with a large shadow economy need not have the largest unemployment rate, confirming what we have just suggested –that a significant part of shadow economy activities are undertaken by those who are not unemployed.

Unemployment Support Programs in the oecd Countries

10The social security system has a number of objectives regarded as necessary to ensure its effectiveness, [5] namely (i) to provide financial support to those who have become unemployed as a result of changes in labour market conditions; (ii) to provide labour market support programs to assist the unemployed back to work; (iii) to provide financial support to those individuals and families in receipt a low income; and (iv) to provide support to those that are unable to work as a result of a sickness or disability. Despite sharing these common set of objectives, the social security systems across the oecd countries vary considerably, not only in terms of their generosity (replacement rates) but also their entitlement criteria (employment) and their method of redistribution.

11These very issues create dilemmas for policy markers entrusted to design and implement an effective social security system (see Adema [2006]). If the level of generosity is too low the social security system fails to maintain adequate support for those experiencing financial hardship while on the other hand, a very generous system may encourage welfare dependency. For the purposes of our discussion, this “dependency trap” may also motivate participation in the shadow economy because the combination of a generous social security system and shadow economy income may provide a level of net income in excess of that received from legitimate employment. In addition, the “dependency trap” may cross generations and the culture of participating in the shadow economy may entrench itself such that participation in the shadow economy is viewed by some as acceptable behaviour.

Comparing Unemployment Programs in the oecd

12In Table 2 we provide an overview of the various social security programs across 28 oecd countries. The social security programs include combinations of Unemployment Insurance (ui), Unemployment Assistance (ua), Social Assistance (sa), housing benefits, family benefits, lone parent benefits and childcare benefits.

Table 2

Oecd Social security Programs

Oecd Country Unemp Insurance Unemp Assistance Social Assistance Housing Benefits Family Benefits Lone Parent Benefits Childcare Benefits Australia – Y Y Y Y Y Y Austria Y Y Y SA Y – Y Belgium Y – Y – Y – Y Canada Y – Y SA Y Y Y Czech Rep Y – Y Y Y – Y Denmark Y – Y Y Y FB Y Finland Y Y Y Y Y FB Y France Y Y Y Y Y Y Y Germany Y Y Y Y Y T Y Greece Y Y – Y Y – – Hungary Y – Y Y Y FB Y Iceland Y – Y Y Y Y – Ireland Y Y Y SA Y Y Y Japan Y – Y SA Y Y – Korea Y – Y SA – Y Y Luxembourg Y – Y SA Y T Y Netherlands Y Y Y Y Y T Y Nz – Y – Y Y Y – Norway Y – Y Y Y Y Y Poland Y – Y Y Y CCB Y Portugal Y Y Y – Y T – Slovak Rep Y – Y Y Y – Y Spain Y Y Y – Y T – Sweden Y Y Y Y Y Y – Switzerland Y – Y SA Y – – Uk Y Y Y Y Y – Y Us Y – Y – Y T Y Information as of 2002. Y indicates that specific benefit or tax credit exists. When no specific housing or lone-parent benefit is available, “sa” (social assistance), “fb” (family benefit) or “ccb” (childcare benefit) indicates that housing or lone-parent specific provisions exist as part of these schemes. (*) – social assistance was at an experimental level and concerned only 305 municipalities (out of more than 8000) in 2002. Source: oecd.

Oecd Social security Programs

13Both ui and ua constitute the principal support mechanism for financially compensating those that become unemployed and are entitled to income support. Although both ui and ua are intended to support the unemployed, there is major underlying difference between the two systems (see World Bank [2004]). Unemployment Assistance (ua) has the principal goal of reducing the extent of poverty, while ui’s objective is to smooth consumption. These principle objectives are also signalled in their eligibility criteria. Essentially ua is a means tested unemployment support scheme while ui is conditional on the claimants demonstrating a previous work history, which is partially self-funded through the payment of premiums. Social Assistance (sa) on the other hand is intended to serve the low-income group (see Vroman and Brusentsev [2005]). The majority of those receiving sa are families with low income and dependents, either children or family members with physical or other medical disabilities. Those receiving sa, who are also unemployed, are expected to be registered for work. Although the ua scheme’s objective is to reduce the extent of poverty, it differs from sa, (although sometimes the distinction is not very clear), in that ua principally provides support for employment reintegration while sa provides a wider range of services including family payments, childcare and other similar types of assistance.

14A glance down column 2 of Table 2 shows that ui is the most common form of financial support for the unemployed. A number of these oecd countries also provide ua assistance along side ui benefits. Eleven of the 28 countries provide both ui and ua while 15 countries provide only a system of ui. Australia and New Zealand are the only two countries listed in Table 2 that do not have an ui system and rely solely on ua for compensating the unemployed. [6] Unemployment assistance in Australia and New Zealand, unlike the ui system, is entirely funded by general taxation unlike ui that relies on employee-employer direct contributions.

15For the purposes of this study we have selected 12 countries of the 28 listed in Table 2, which appear shaded. These countries have been chosen based on the various combinations of social security support outlined in Table 2. Of these 12 countries, two offer only ua (Australia and New Zealand), five offer only ui (Belgium, Canada, Denmark, Italy and Norway) and five offer both ui and ua (Finland, France, Portugal, Sweden and the uk). In addition, the five countries that offer only ui support have varying combinations of social security programs (e.g. housing benefits, family benefits, lone parent benefits and childcare benefits) as do the other five countries that provide both ui and ua support.

16Of the various objectives of this paper, one of these is to examine whether the types of (unemployment) social security programs affect differently the size of shadow economy participation. In the process of drawing conclusions, various factors such as eligibility criteria, generosity level and duration periods will be considered as these ultimately determine the overall generosity and the ease by which claimants can access these benefits. It is to these three specific issues that we now turn our attention.

17Whether the claimant is applying for ui or ua support, the eligibility criteria, duration and benefit payments vary significantly across the oecd countries. In Tables 3 and 4 we provide the eligibility criteria, waiting period, duration and payment benefits for ui and ua schemes respectively in each of the 12 countries that we have identified in Table 2. In Table 3 we present an overview of the various employment conditions, waiting periods, maximum duration and benefit payments for ui assistance.

18Column 2 of Table 3 documents the employment criteria for ui eligibility in each of the chosen oecd countries, excluding Australia and New Zealand that do not provide an ui scheme. A glance down column 2 identifies the variability in the “employment condition” eligibility criteria. For example, in Portugal a claimant is required to work a period of 540 days over the previous 2 years (or approximately ¾ of the time in employment), while in Canada a period of 630 hours of employment over the previous year is the expected minimum requirement (or approximately 1/3 of the time in employment). Finland requires a period of 34 weeks employment over a 2 years period while in Norway there is no specific employment criterion. Of these 10 countries that provide ui insurance programs, only Denmark, Finland and Sweden, have a voluntary system of contributions. In the remaining 7 countries, ui insurance is compulsory during a period of employment.

Table 3

Unemployment Insurance Programs

Oecd Country Employment Conditions Voluntary Compulsory for employees Waiting Period (days) Max Duration (months) Payment Rate (% of base earnings) Min Benefit (% of apw) Max Benefit (% of apw) Belgium 468 days in 27 months C 0 Unlimited 60 (50 after 1 yr) 28 (25) 39 Canada 630 hours in 1 year C 14 9 55 – 55 Denmark 52 weeks in 3 years V 0 48 90 Depends on employment record 52 Finland 34 weeks in 2 years* V 7 23 (21% of apw) + 45% of earnings exceeding basic benefit – – France 4 months in 18 months C 8 30 57-75 40 295 Italy 52 weeks in 2 years C 7 6 40 – 52 Norway – C 3 36 62 19 111 Portugal 540 days in 2 years C 0 24 65 50 149 Sweden 6 months in last year V 5 14 80 35 (50) 76 (70) Uk 2 years C 3 6 Fixed at 14% of apw – – Data shown for 2002, for a 40-year-old single worker without children, with a 22-year employment record. Source: oecd.

Unemployment Insurance Programs

19The maximum entitlement and benefit payments also vary considerably across each of the 10 countries. For example, the maximum duration period for ui assistance in the Italy, Finland, Norway and Denmark are 6, 23, 36 and 48 months respectively. In Belgium there is no maximum duration period. The benefit payment rates in Italy, Finland, Norway and Denmark are 40% (of payment rate), 21% of the Average Production Wage (apw) (plus 45% of earnings exceeding the basic benefit), 62% (of base earnings) and 80% (of base earnings) respectively. Denmark appears the most generous of all the 10 countries in Table 3, with the highest payment rate (as a percentage of base earnings) while in Belgium, Canada and Italy the payment rates are markedly different at 60% (50% after 1 year), 55% and 40% respectively. The payment rate in the uk is the lowest, fixed at 14% of the apw. We will return to these payment rates in subsequent sections of this paper to determine whether a correlation exists between the size of the shadow economy and the generosity of the social security programs. This in part might suggest that participation in the shadow economy by the unemployed may be prolonged by the generosity of the social welfare system thereby increasing the size of the shadow economy.

20When an unemployed individual has exhausted their entitlements for ui benefits, they may also be eligible for ua benefits. In Table 4 we provide an overview of the employment conditions, waiting period, duration and benefit payments entitlements for ua assistance.

Table 4

Unemployment Assistance

Oecd Country Employment Record (months) Waiting Period (days) Max Duration (months) Payment Rate (% of base earnings) Max Benefit (% of apw) Australia – 7 unlimited Fixed amount 20 Finland – 5 unlimited Fixed amount 21 France UI and 60 in last 120 – 6 months (renewable) – 22 Nz – 7-70 unlimited Fixed amount 24 Portugal UI or 6 in last 12* – 12 (after ui) or 24 Fixed amount 40 Sweden 6 or recent graduate 5 14 Fixed amount 35 (30) Uk – - unlimited Fixed amount 14 Data shown for 2002, for a 40-year-old single worker without children, with a 22-year employment record. ui: unemployment insurance; ua: unemployment assistance. Source: oecd.

Unemployment Assistance

21Four of the 7 countries listed in Table 4, namely Australia, Finland, New Zealand and the uk, provide ua benefits without any employment conditions while France Portugal and Sweden require some form of employment record. There are no limits on duration for ua support in Australia, Finland, New Zealand and the uk, although the waiting period in New Zealand may be as much as 70 days. The maximum duration for France, Portugal and Sweden varies between 6 and 24 months depending on individual circumstances.

22The various eligibility criteria and payment benefits detailed in Tables 3 and 4 ultimately determine how many people receive support under the ui and ua schemes. It is quite possible that fewer than the total number of people unemployed will receive ui or ua support because they may not meet certain eligibility criteria. Therefore it is possible that participation in the shadow economy may be made up in part by those unemployed receiving assistance and in part by those that are not.

Replacement Rates and the Generosity Index

23A contributing factor prolonging the duration of unemployment is the level of benefit generosity. Combining this with the ease by which an unemployed person may participate in the shadow economy compounds this spell of unemployment duration. A generous social security system could potentially provide a level of income support (after including shadow economy income) above that which legitime employment could provide, net of taxes and other contributions. Of course much of these shadow employment activities are likely to be concentrated in specific occupations, for example, those that may not necessarily command a high wage in the legitimate economy and/or the service sector (for example, the services of a hairdresser).

24Constructing measures of the replacement rates and generosity indices for each country provides a way of comparing the level of generosity of the social security systems across countries. The replacement rate is the defined as the ratio of unemployment benefits received as a proportion of the worker’s earnings. Calculating this ratio before adjusting for family benefits, childcare payments, taxation and other social assistance provides a measure of the gross replacement rate. Adjusting for taxation, social security contributions and the other forms of social assistance provides a measure of the net replacement rate. The net replacement rate thus allows for a more effective comparison of the social security systems because it considers each country’s specific taxation rulings on social security payments and other forms of tax credits.

25Nevertheless, constructing the replacement rates for each country is fraught with complexity because social security entitlements (ua and sa) are paid conditional on individual and family circumstances – for example, whether a person is single or married and if married, whether or not they have children (and how many). The oecd publishes gross and net replacement rates for various family types and levels of earnings. For the purposes of this paper we use the unweighted average of the various replacement rates as a means for simplification. Consequently these rates are calculated in the relation to the household as a whole and not by individual circumstances.

26The generosity index, G, is a convenient metric to account not only for the replacement rate but also the recipiency rate. [7] The generosity index is calculated as follows:

27

28where

29RRate: replacement rate; and

30NBen/Unemp: recipiency rate (= the ratio of the average weekly number of beneficiaries to the average weekly number of unemployed).

31The value of G can vary through government policy changes that affect either the replacement rate and/or the recipiency rate. It is important to note that the number of people receiving support as calculated in the recipiency ratio, (NBen/Unemp), is not necessary nested within the denominator (the number of unemployed) because both ui and ua recipients are included in the measure of the numerator. For example, Vroman and Brusentsev [2001] find that from 1995 to 1999 that the recipiency rate in Australia exceeded a value of 1. [8] During the 1990’s a similarly larger value of the recipiency rate was found for Denmark, Ireland, Netherlands and New Zealand (see Vroman and Brusentsev [2005] for details).

32In Figure 1 we plot the size of the shadow economy against the generosity index, calculated using the gross replacement rates, for the year 2002. Figure 1 suggests that a positive relationship between the size of the shadow economy and the generosity index exists. This positive relationship is more obvious if we exclude Portugal from the sample, which appears to be an outlier.

Figure 1

The Shadow Economy and the Generosity Index for 2002

The Shadow Economy and the Generosity Index for 2002

33This figure might at first imply, without further investigation, that a more generous social security system may help prolong unemployment duration, and by contributing to subterranean activity, increases the overall size of the shadow economy. Interestingly however, Portugal has a comparatively low generosity index (G), yet the size of its shadow economy is quite large when compared those countries with similar levels of G (Australia and the uk). [9]

34In Figure 2 we examine how the relationship between the size of the shadow economy and the generosity index has changed. Using data from Vroman and Brusentsev [2005] we compare the change between the 1990’s and 2002. [10] The arrows in Figure 1 indicate the direction of change.

Figure 2

Comparing Changes in the Generosity Index and the Shadow Economy

Comparing Changes in the Generosity Index and the Shadow Economy

35If increases in the generosity of social security payments induce prolonged unemployment and participation by the unemployed in the shadow economy, we might expect the arrows in Figure 2 to point in the north-easterly direction. This appears to be the case for Denmark, the uk and New Zealand which each exhibited increases in their generosity index and the size of their shadow economies. For example, during the 1990’s the generosity index for Denmark averaged 0.53 while its shadow economy averaged 15.9% of gdp. By 2002 the generosity index climbed to 0.63 and the size of the shadow economy was estimated at 17.6% of gdp. For Denmark the increase in the generosity index was primarily generated by increases in the recipiency rate, while in New Zealand and the uk the increase in the generosity index was the culmination of increases in both the replacement and recipiency rate, with the latter being more prominent. In Denmark, the uk and New Zealand, the higher generosity index in 2002 also coincided with a larger estimate of the size of the shadow economy.

36When we consider instances of declining generosity indices, there are two valid but alternative scenarios on the effects of shadow economy participation by the unemployed. First, declining generosity might reduce unemployment duration spells and reduce unemployment participation in the shadow economy, thereby reducing its size. Second, declining generosity might induce a more active participation in the shadow economy by those currently unemployed, thereby increasing its size. The results in Figure 2 appear to conform to the second scenario, that reductions in the size of the generosity index leads to increases in the size of the shadow economy. The generosity index in Figure 2 for Australia, France and Portugal declined between the 1990’s and 2002, but only moderately for Australia. In each of these three countries the size of the shadow economy increased over this time. When we consider how the replacement and recipiency rates contributed to the decline in the generosity index we find that for each of these three countries the decline was explained solely by falls in the replacement rates. In fact the recipiency rate increased in each of the three countries (by approximately 20% in Australia and Portugal and by approximately 7% in France). So it seems at a first glance that the decline in the generosity index, generated by a decline in the replacement rate, may have contributed to the increase in the shadow economy for these countries. [11]

37In Figure 2 we re-examine this relationship using instead the net replacement rate to calculate the generosity index. In Figure 2 the relationship between the generosity index and the size of the shadow economy is more pronounced than in Figure 1. [12] This is expected because the decision to prolong unemployment and work in the shadow economy is driven primarily by income motivations, or more precisely, by comparisons of net income.

38In Australia, France and New Zealand, the lower generosity index coincides with a relatively smaller shadow economy, whereas in Denmark, Norway, Finland, Sweden and Belgium a higher generosity index coincides with a larger estimate of the shadow economy. Portugal, and to a lesser degree the uk, provide an exception to the other countries. In the uk the comparatively higher generosity index coincides with a small shadow economy while in Portugal, a large shadow economy coincides with a very low generosity index.

Figure 3

The Shadow Economy and the Generosity Index Using nnr: 2002

The Shadow Economy and the Generosity Index Using nnr: 2002

The Income and Substitution Effects in Shadow Economy Activity

39The inferences drawn from the previous figures are complicated by the fact that the shadow economy comprises of two effects –what we will call the “income effect” and the “substitution effect”. Each of these effects operate in opposite directions, producing an overall net effect (the size of the shadow economy) that we plotted on the vertical axis in Figures 1, 2 and 3. Very briefly, the income effect is a measure of shadow economy participation that is driven by opportunities arising from an expanding legitimate economy, while the substitution effect is a measure of shadow economy participation by those unemployed who are seeking to smooth their consumption expenditures along with the support of government financial assistance (social security payments), if available. The substitution effect typically increases during declining periods of legitimate economic activity as a result of unemployment, while at the same time the size of the shadow economy declines as a result of overall declines in economic activity (the income effect). That is, during the course of a business cycle, the income and substitution effects move in opposing directions and our measures of the shadow economy capture the net result of these effects. In this section we provide the first preliminary attempt at gauging the extent and characteristics of the substitution effect component of the shadow economy in the oecd as a way of motivating research in this area. We will specifically examine how this relates to changes in unemployment and whether differences in social security programs impact on shadow economy participation by the unemployed.

40There have been a number of studies that have examined the business cycle properties of the shadow economy (see Giles [1997a], [1997b], [1999]; Giles and Tedds [2002]; Bajada [2003]; and Busato and Chiarini [2004]). These empirical findings suggest that a procyclical relationship exist between the legitimate and shadow economies. This makes intuitive sense if one accepts that opportunities in the shadow economy are dependent on the opportunities available in the legitimate economy. For example, a tradesman is likely to engage in substantial shadow economy activity when the legitimate economy (and the construction sector) is booming. During declining periods of legitimate activity (including declines in the construction sector), tradesmen will find themselves in the situation of having less legitimate work, of which an even smaller proportion than before may go unreported (the shadow economy). Thus the growth in the legitimate economy is the stimulus for providing opportunities in the shadow economy. The findings of this earlier research tend to support this argument.

41In Figure 4 we plot the growth cycles of the shadow and legitimate economy using the Hodrick-Prescott [1997] (HP) filter. [13] We observe from Figure 4 that a strong procyclical relationship exists between the legitimate and shadow economy growth cycles. [14]

Figure 4

Growth Cycles of the Shadow and Legitimate Economies

Growth Cycles of the Shadow and Legitimate Economies

42Despite these procyclical relationships evident in Figure 4, we have argued that the growth of the shadow economy comprises of two opposing forces –the “income effect” and the “substitution effect”. The following example highlights this point. As incomes decline we would expect to see an overall decline in the consumption of goods and services –both in the legitimate and the shadow economy. This is the “income effect”. At the same time however, rising levels of unemployment generally follow declines in legitimate economic activity as businesses lay-off workers to compensate for falling demand. If those who become unemployed are enticed to work in the shadow economy to supplement any government assistance they might receive, we should expect to observe rising levels of shadow economy activity. [15] This is the “substitution effect”. Whichever effects dominates will determine whether the shadow economy moves procyclically or counter-cyclically with the legitimate economy. If the “income effect” dominates we would observe a positive correlation between the growth cycles of the two sectors and conversely if the substitution effect dominates. The growth cycles of the legitimate and shadow economy in Figure 4 clearly suggest that the “income effect” strongly dominates the “substitution effect” producing the procyclical business cycle relationships that we observe.

Estimating the Substitution Effect

43As we have suggested, the results of Figure 4 imply the income effect strongly dominates the substitution effect. In fact we would expect to find an even stronger correlation between the growth cycles of the legitimate economy and the corresponding “income effect”, if the income effect were directly observable. However it is nevertheless possible with these general results to make a first attempt at estimating the cyclical component of the substitution effect. To briefly explain the method by which we estimate this “substitution effect”, we begin with a simple definition of the growth rate of the shadow economy (equation 2) and the identity that changes in the size of the shadow economy is the result of changes in the income and substitution effect components separately (equation 3) which we use here for illustrative purposes. As we have already suggested, the income and substitution effects move in opposing directions, so when one measure is positive (say, the income effect) the other measure (the substitution effect) is negative. [16]

44

45where U^I = the income effect and U^S = the substitution effect substituting (3) into (2) gives

46

47If, for the moment, we assume that: (i) there is no substitution effect (?U^S_t = 0); (ii) the business cycles of the legitimate and shadow economy are strongly procyclical (which we have observed in Figure 4); and (iii) an individual conceals a constant proportion of legitimate income over the course of the business cycle such that the growth rate in legitimate income is mirrored by the growth in shadow economy income (the income effect component), we can write:

48

49where Y = gdp.

Figure 5

The Relationship between the Cyclical Component of the Substitution Effect and the Change in Unemployment (1993-2005)

The Relationship between the Cyclical Component of the Substitution Effect and the Change in Unemployment (1993-2005)

50However there are many activities in the legitimate economy as measured by gdp, such as industrial mining and the production of passenger vehicles and heavy industrial equipment that are unlikely to be produced in the shadow economy. Thus, using a measure of the growth rate of gdp may not be adequately representative of the shadow economy cycles or the “income effect”. From the various industry studies of shadow economy participation (see cetf [1998], [2003]; Statistics Canada [1994]), the sectors typically identified as contributing substantially to the shadow economy include the services, construction and agricultural sectors. Therefore as a way of improving on the approximation given by equation (5), we replace the use of gdp (= Y) with an output measure of the services sector (Y^S). [17] Assuming the substitution effect is negative (or declining), replacing Y with Y^S in equation (5) and substituting this into equation (4) gives:

51

52which by rearranging, gives an approximation for the growth share of the “substitution effect”:

53

54In Figure 5 we plot the growth cycle of the “substitution effect” [18] against the change in the unemployment rate for each of the 12 oecd countries, whilst in Figure 6 we plot the gdp growth cycle against the change in the unemployment rate (Okun’s law).

55The results of Figure 6 illustrate a similar Okun-style relationship of the effects of changes in unemployment on the “substitution effect” of the shadow economy. As expected the relationship is positive, confirming our earlier hypothesis from Figures 1, 2 and 3 that the unemployed are engaging in shadow economy activity. Although the positive relationship is clear across the 12 countries, the strength of the relationship illustrated in Figure 5 is limited by the short time span covered in this study. We expect that by increasing the length of the time series estimate of the shadow economy we would observe a much stronger relationship between the change in the unemployment rate and the estimate of the “substitution effect”. This observation is prompted by the following observations: (i) the Okun relationship in Figure 6 estimated over the same time period defined in Figure 5 is much weaker than typically observed over much longer time duration; and (ii) the strength of the relationship in Figure 5 resembles closely the strength of the relationship in Figure 6.

Figure 6

Okun’s Law (1993-2005)

Okun’s Law (1993-2005)

56In Table 5 we present, for the purposes of comparison, the estimates of the Okun coefficient for the period 1993-2005 [column (i)] with Lee’s [2000] estimates of the Okun coefficients for the period 1955-1996 using various filters to extract the cyclical component [column (ii)].

Table 5

Estimates of the Okun and the Substitution Effect Coefficients

Okun’s coefficient (1993-2005) Okun’s coefficient Lee (2000) (1955-1996) Substitution effect coefficient (1993-2005) (i) (ii) (iii) Australia – 1.67 – 0.85 to – 1.81 + 1.77 Belgium – 0.97 – 1.03 to – 1.12 + 0.70 Canada – 0.11 – + 1.02 Denmark – 0.90 – 0.99 to – 1.30 + 1.49 Finland – 1.21 – 1.44 to – 1.73 + 0.61 France – 1.73 – 2.20 to – 2.91 + 0.69 Italy – 1.77 – 0.57 to – 2.41 + 1.47 New Zealand – 0.62 – + 1.35 Norway – 0.18 – 2.21 to – 3.84 + 0.62 Portugal – 0.27 – + 0.46 Sweden – 1.25 – 1.54 to – 1.87 + 0.98 UK – 1.03 – 1.39 to – 1.51 + 1.47

Estimates of the Okun and the Substitution Effect Coefficients

57Our estimates of the Okun coefficients for Australia and Italy fall within the range of estimates calculated by Lee [2000] while for the remaining countries, our estimates of the Okun coefficients fall outside this range, reinforcing point (i) in the preceding paragraph. [19] The “substitution effect” coefficients are given in column (iii) of Table 5. In Australia the results suggest that a 1 percentage change in the unemployment rate increases the substitution effect component of the shadow economy by 1.77% while in Portugal the effect is much lower (0.46%).

Unemployment Benefit Programs and Shadow Economy Behaviour

58Do the various unemployment programs influence the extent of shadow economy participation? We have already found varying degrees of shadow economy participation by the unemployed in the results illustrated in Figure 5. How this is influenced by the various unemployment programs and employment support mechanisms is not clear however by a simple inspection of these visual plots. In Figure 7 we plot the correlation between the change in the unemployment rate and the growth cycle of the “substitution effect” (on the vertical axis) against the replacement rates (on the horizontal axis) calculated for the period 2000-2003. [20]

Figure 7

Relationship between the Replacement Rate and the Correlation between the Substitution Effect and Unemployment (2002)

Relationship between the Replacement Rate and the Correlation between the Substitution Effect and Unemployment (2002)

59From Figure 7, a significant negative relationship is evident between correlation and the replacement rate. The data in this figure suggests that countries with low replacement rates exhibit higher correlations between the “substitution effect” and the change in the unemployment rate. On the other hand, in those countries with larger replacement rates, the correlations are much lower. In the same figure there appears to be two notable groupings of countries – one group of 4 countries (Australia, New Zealand, the uk and Sweden) with replacement rates less than 0.3 and correlations of 0.3 and above and a second group of countries (France, Finland, Belgium, Norway, Portugal and Denmark) with replacement rates greater than 0.3 and correlations of less than 0.3. Although there is no significant reason for choosing a correlation and a replacement rate of 0.3, the two identified clusters of countries each share very common set of characteristics. We have defined each cluster as Region I (for those countries with a replacement rate less than 0.3) and Region II (for those countries with a replacement rate greater than 0.3). In Table 6 we provide summary statistics on the average size of the shadow economy, the average generosity index [21] and the average “substitution effect” coefficient implied from Figure 5.

Table 6

Summary Statistics: Regions I and II

Region I Region II Ave size of Shadow Economy (% of gdp) 14.5% 18.97% Ave Generosity Index 0.32 0.52 Ave Okun “substitution effect” coefficient 1.38 0.76

Summary Statistics: Regions I and II

60The results in Table 6 present a number of interesting findings. Those countries in Region I have a relatively smaller shadow economy and share a typically larger “substitution effect” coefficient when compared with those countries in Region II. The larger correlation coefficients for those countries in Region I may be the consequence of the smaller generosity index, which is primarily driven by low replacement rates (as shown in Figure 7). Although there are some variations in the recipiency rates across the 10 countries, the differences are small. Therefore it is possible, for those countries that have a low correlation coefficient, that the a larger replacement rate may be encouraging longer duration spells in unemployment. That is, those who become unemployed and engage in the shadow economy while in receipt of social security payments in countries with higher replacement rates (Region II) may tend to exhibit longer unemployment duration spells than in countries with lower replacement rates (Region I). [22] In fact this conforms to the empirical literature (and theoretical expectations) which find that not only do unemployment benefits encourage extended durations in unemployment, but the more generous these payments, the longer is the duration. [23] This might explain the smaller correlation coefficients in Figure 7 for those countries that have higher replacement rates.

61We examine the possibility of longer unemployment duration spells in Figure 8 by examining how the correlation in period (t + 1) differs from the correlation in period (t) against the same change in the unemployment rate in period (t). These two correlation-unemployment combinations are identified by “x” and “o” respectively in Figure 8.

Figure 8

Relationship between the Replacement Rate and the Correlation between the Substitution Effect and Unemployment – Lag one period

Relationship between the Replacement Rate and the Correlation between the Substitution Effect and Unemployment – Lag one period

62For those countries in Region II, the positive correlation coefficient between the cyclical components of the substitution effect in period (t + 1) and the change in unemployment in period (t), increases. [24] This implies that a change in the unemployment rate today is related more strongly with subsequent changes in the substitution effect in period (t + 1). How far into the future the correlations continue to increase must at the very least be bounded by the maximum duration for benefit entitlements, which for some countries is quite short and for others is quite substantial. In Table 7 we present the cross-correlations coefficients between the substitution effect [from period (t ? 6) through to period (t + 6)] and the change in unemployment in period (t).

63The cross-correlations in Table 7 suggests that the cyclical components of the substitution effect are strongly procyclical (with a lead) in Belgium, Canada, Denmark, Finland, France, Norway, Portugal and the uk. The lead varies across countries, with Canada and Denmark showing the strongest correlation with a lead of 3 quarters, Belgium France, Portugal and the uk with a lead of 2 quarters, and Finland and Norway with a lead of 1 quarter. [25] The contemporaneous correlation in Australia, Italy, New Zealand and Sweden were found to be the largest suggesting that any effect on unemployment duration may occur within a period of 3 months. [26]

64We may interpret the timing of the maximum correlation coefficients in Table 7 as the delayed duration of unemployment influenced by the combination of a generous welfare system and participation in the shadow economy. [27] These lags are transcribed in months in column (xiv) in Table 7. In column (xv) of the same table, the maximum duration periods of unemployment benefit entitlements from Table 3 are presented. Comparing columns (xiv) and (xv) we find that for those countries were unemployment is delayed, the average duration spells is approximately 6 months. Only in Canada and the uk does it appear that unemployed recipients working in the shadow economy exhaust their unemployment entitlements. In Norway, France and Denmark where maximum duration of unemployment benefits exceed 2 years, the average delay in the duration of unemployment varies between 3 months (for Norway) to 9 months (for Denmark). In Belgium where the duration of entitlements is unlimited, the delay in unemployment duration is approximately 6 months. In Australia where ua is unlimited in duration, the delayed duration in unemployment by those participating in the shadow economy appears to be less than 3 months. [28]

Table 7

Cross Correlation Between the Cyclical Component of the Substitution Effect and Unemployment

The Substitution Effect with Lead or Lag relative to ?ue Procyclical with lead (months) Max ui Duration (months) – 6 – 5 – 4 – 3 – 2 – 1 0 + 1 + 2 + 3 + 4 + 5 + 6 (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) (x) (xi) (xii) (xiii) (xiv) (xv) Australia 0.025 0.065 0.120 0.177 0.271 0.392 0.535 0.424 0.332 0.263 0.212 0.219 0.115 0 – Belgium 0.056 0.034 0.011 0.059 0.109 0.216 0.214 0.302 0.463 0.461 0.376 0.433 0.351 6 Unlimited Canada 0.099 0.039 0.077 0.110 0.112 0.176 0.223 0.389 0.432 0.438 0.382 0.178 0.144 9 9 Denmark 0.131 0.092 – 0.028 – 0.087 – 0.086 0.049 0.246 0.442 0.612 0.664 0.593 0.447 0.263 9 48 Finland – 0.097 – 0.234 – 0.259 – 0.240 – 0.327 – 0.145 0.078 0.253 0.178 0.187 0.048 – 0.021 – 0.014 3 23 France – 0.161 – 0.233 – 0.215 – 0.147 – 0.020 0.112 0.232 0.292 0.293 0.239 0.162 0.059 – 0.058 6 30 Italy 0.143 0.189 0.185 0.258 0.298 0.310 0.402 0.347 0.298 0.209 0.108 0.075 0.022 0 6 New Zealand 0.029 0.027 0.017 0.041 0.152 0.182 0.216 0.164 0.058 0.064 0.046 0.102 0.154 0 – Norway 0.137 – 0.006 0.052 – 0.096 0.054 – 0.017 0.048 0.223 0.064 0.089 – 0.055 – 0.084 0.013 3 36 Portugal 0.201 0.115 0.136 0.183 0.004 0.121 0.192 0.304 0.392 0.253 0.310 0.172 0.159 6 24 Sweden 0.007 0.012 – 0.066 – 0.049 – 0.052 0.165 0.338 0.255 0.162 0.035 0.003 0.001 0.050 0 14 Uk 0.143 0.184 0.105 0.066 0.103 0.156 0.254 0.325 0.344 0.283 0.232 0.208 0.112 6 6

Cross Correlation Between the Cyclical Component of the Substitution Effect and Unemployment

65In Table 8 we present the results of estimating a model to explain cyclical variations in the unemployment rate. In this specification we have included: (i) the cyclical variation in economic activity (gdp) which is intended to capture the negative relationship we know as Okun’s law; (ii) the cyclical variations in real social security payments made by government to capture any substantial changes in welfare payments that may affect changes in unemployment [29]; (iii) the cyclical variation in the real wage rate; and (iv) the degree of openness of the economy. We also include in the model leads of the “substitution effect” in order to test the significance of this variable after controlling for the other factors influencing the changes in unemployment. [30] We expect that the sign on the “substitution effect” coefficient to be positive. To measure the significance of the relationship between the substitution effect and the unemployment rate, we estimate the following equation: [31]

66

67where

68UE^cyc = the growth cycle of the unemployment rate;

69Y^cyc = the growth cycle of gdp;

70SocBen^cyc = the growth cycle of social security benefit payments paid by government;

71Wage^cyc = the growth cycle of the real wage (using compensation of employees);

72U^S_t+i = the cyclical component of the substitution effect of the shadow economy, where t + i defined as the lead that is determined by the maximum cross-correlation coefficients in Table 7.

73Open^cyc = growth cycle of the measure of openness, calculated as the sum of exports and imports as a percentage of gdp.

74The results in Table 8 suggest that cyclical variations in economic activity (as measured by the growth cycle of gdp) contributes significantly towards explaining changes in unemployment while the coefficients on social security payments by government and the real wage are generally weak in explaining the short-run fluctuations in the unemployment rate. For those countries where the degree of openness was found to be significant (Canada, France and Sweden), the sign on the coefficient is negative as expected (see Romer [1989]). What we find most interesting from these results is that the “substitution effect” coefficient is not only positive for the 12 oecd countries but is also significant for 10 of the 12. Only for Sweden and Italy do we find an insignificant coefficient on the “substitution effect”, although the coefficient is positive as expected. These results of Table 8 would imply that changes in unemployment do in fact affect participation in the shadow economy, but the effects are most likely to be transitory rather than permanent.

Table 8

Results of Estimating the Effects on the Cyclical Changes in Unemployment

Constant ?0 ? 1 ? 2 ? 3 ? 4 ? 5 ? 6 Adj R2 Australia 0.001 0.756 – 1.572 0.045 – 0.402 1.026 – 0.89 (0.25) (13.12)** (– 2.79)** (0.63) (– 1.22) (2.85)** – Belgium 0.002 0.645 – 2.255 0.214 – 0.497 1.012 – 0.81 (0.37) (6.76)** (– 3.46)** (0.28) (– 0.96) (2.20)** – Canada – 0.0004 0.799 – 1.841 – 0.276 1.211 1.116 – 0.75 (– 0.08) (5.04)** (– 1.68) (– 0.70) (1.26) (1.83)* – – 0.002 0.317 – 2.053 – 0.640 1.478 0.882 – 1.162 0.82 (– 0.38) (1.66) (– 2.23)** (– 1.85)* (1.82)* (1.71)* (– 3.51)** Denmark 0.005 0.685 – 2.709 0.455 0.263 1.310 – 0.80 (0.98) (7.63)** (– 3.57)** (1.50) (0.36) (3.81)** – Finland 0.002 0.712 – 1.987 – 0.148 0.279 0.996 – 0.90 (0.45) (6.63)** (– 2.74)** (– 0.429) (0.97) (1.82)* – France 0.003 0.779 – 2.369 – 0.297 0.859 0.655 – 0.95 (1.80)* (10.81)** (– 6.78)** (– 0.76) (2.10)** (4.69)** – 0.003 0.683 – 1.531 – 0.094 0.119 0.628 – 0.285 0.95 (1.92) (7.40)** (– 3.05)** (– 0.24) (0.23) (4.67)** (– 2.24)** Italy 0.0007 0.635 – 0.859 0.170 – 0.601 0.132 – 0.82 (0.35) (7.28)** (– 2.82)** (0.69) (– 2.24)** (0.46) – New Zealand 0.0008 0.623 – 0.793 1.085 – 0.007 1.237 – 0.77 (0.15) (4.96)** (– 1.16) (1.62) (– 0.01) (2.38)** – Norway 0.0002 0.566 – 0.025 0.888 – 1.628 1.183 – 0.61 (0.03) (4.07)** (– 0.03) (1.42) (– 1.86)* (1.70)* – Portugal 0.014 0.467 – 4.290 – 2.058 – 2.456 4.705 – 0.79 (1.34) (3.22)** (– 2.18)** (– 2.20)** (– 1.77)* (3.78)** – Sweden – 0.003 0.678 – 2.663 – 0.201 – 0.187 0.597 – 0.83 (– 0.40) (7.22)** (– 2.92)** (– 0.39) (– 0.35) (0.78) – – 0.002 0.766 – 0.913 – 0.050 – 0.041 0.585 – 1.029 0.86 (– 0.26) (8.57)** (– 0.93) (– 0.11) (– 0.08) (0.84) (– 3.29)** Uk – 5.1E-05 0.621 – 1.688 0.256 0.009 0.831 – 0.85 (– 0.02) (7.33)** (– 3.30)** (1.54) (0.03) (2.06)** – (*) indicates significance at 10% and (**) indicates significance at 5%. Values in parenthesis are t-statistics.

Results of Estimating the Effects on the Cyclical Changes in Unemployment

75Does the type of unemployment support program differ in its effect on shadow economy activity? Does it matter whether a country provides only ui, ua or both on the level of participation in the shadow economy? If we consider the maximum correlations given in Table 7, we observe that the correlations across each country are in fact relatively similar, with the exception of Denmark. In Figure 9 we plot these maximum correlation coefficients against the generosity index calculated using the net replacement rate.

Figure 9

The Maximum Correlation Coefficient and the Generosity Index

The Maximum Correlation Coefficient and the Generosity Index

76An examination of Figure 9 suggests that there is no relationship evident between generosity, the type of unemployment assistance program, and the correlation between the change in unemployment and the growth cycle of the substitution effect. In fact the correlation coefficient (except for Denmark) ranges between 0.3 and 0.4 while the generosity indices range from 0.49 to 1.1. Portugal and Australia both have a similar generosity index and correlation coefficient despite their shadow economies varying considerably (Australia: 13.9% of gdp, and Portugal: 22.4% of gdp). On the other hand Portugal and Belgium share similar correlation coefficients and estimates of the shadow economy (as a % of gdp), but Belgium’s generosity index is more than double that of Portugal.

77When we compare the type of unemployment assistance programs in each country we find that there is no systematic pattern either. Both Portugal and the uk provide ui and ua support and both have significantly different generosity indices, but the correlation coefficient values are quite similar. On the other hand New Zealand and France have similar generosity indices and correlation coefficients, yet both countries differ in the type of support they provide to their unemployed.

Testing for Asymmetries in the Substitution Effect Component

78Models measuring asymmetric responses have been used relatively extensively in the literature (see Tong [1990]; Terasvirta [1990]; Cover [1992]; Domain and Louton, [1995]; Silvapulle and Silvapulle [1999]; and Bajada [2005]). In this section we consider similar asymmetric (threshold) models to determine whether participation in the shadow economy changes symmetrically or asymmetrically with changes in unemployment. To measure the extent by which increases or decreases (measured separately) in the number of unemployed affect shadow economy activity we estimate the following threshold model, where the threshold level is set to zero:

79

80where

81

82and

83

84The presence of symmetry in equation (9) is evident when

85

86We can apply equation (9) to test whether increases (decreases) in the unemployment rate have steeper effects than do decreases (increases) in the cyclical component of the substitution effect. The appropriate hypothesis is H₀: ?_j = ?_j against the alternative H_a: ?_j ? ?_j for j = 1, 2, …q. If we are interested in knowing whether negative (positive) shocks in the rate of unemployment have steeper effects than do positive (negative) shocks we substitute ?_j = ?_j + ?_j(?_j = ?_j + ?_j) into equation (9), which reduces to equations (11) and (12) respectively:

87

88The null hypothesis for equations (11) and (12) now becomes H₀: ?₁ = ?₂ = … = ?_q = 0. Rejecting the null in each case implies that reductions (increases) in unemployment generate similar downturns (upswings) in shadow economy activity as measured by the substitution effect component.

89In Table 9 we present the results of estimating the asymmetric threshold model of equation (9) for those countries that exhibited evidence of delayed unemployment duration in Table 7. [32] In column (i) of Table 9 are the optimal lag lengths and in columns (ii), (iii) and (iv) are the results for ?UE⁺ (equation 12). The ?² statistics and the p-values in columns (iii) and (iv) suggest that we cannot reject the null hypothesis that the coefficients on ?UE⁺ are jointly zero. In columns (v), (vi) and (vii) we present the results for ?UE^– (equation 11) and we find that the conclusions for negative changes in unemployed are similar to those of ?UE⁺. The results of these tests appear to indicate evidence of symmetry rather than asymmetry on the substitution effect component of the shadow economy from changes in the unemployment rate.

Table 9

Substitution Effect Cycle Responses to Changes in Unemployment from Symmetric and Asymmetric Response Models

Lag length Ue+ Ue– Coefficient Sum ?2 p-val Coefficient Sum ? 2 p-val Belgium (2,2) 0.0130 4.45 0.11 0.012 5.11 0.08 Denmark (3,3) 0.0380 4.33 0.23 0.024 4.55 0.21 Finland (3,1) – 0.0004 0.03 0.85 0.004 2.76 0.10 France (1,2) – 0.0020 3.51 0.17 0.009 3.15 0.21 Norway (1,1) 0.0050 0.33 0.57 0.018 4.35 0.04 Portugal (2,2) – 0.0020 0.71 0.70 0.016 8.18 0.02 Uk (1,2) 0.0020 2.87 0.24 0.007 1.70 0.43 Nz includes the contemporaneous effect.

Substitution Effect Cycle Responses to Changes in Unemployment from Symmetric and Asymmetric Response Models

90How does the magnitude of an increase in unemployment compare to an expansion in shadow economy activity (via the substitution effect)? In Table 10 we present the results of the hypothesis that increases in the unemployment rate produce just as steep an effect as expansions of shadow economy activity. From the p-values, which are all greater than 0.05, we conclude that an increase (decrease) in the unemployment rate have as steep an effect as declines (increases) on the growth cycle of the shadow economy.

Table 10

Changes in Unemployment and Steepness Effects on Shadow Economy Activity

? 2 p-val ? 2 p-val Belgium 2.31 0.31 Australia* 0.02 0.88 Denmark 1.20 0.75 Italy* 0.32 0.57 Finland 1.51 0.22 Sweden* 0.07 0.80 France 1.03 0.60 Nz* 0.002 0.96 Norway 0.67 0.41 Portugal 3.65 0.16 Uk 1.47 0.48 (*) includes contemporaneous effect.

Changes in Unemployment and Steepness Effects on Shadow Economy Activity

91We can provide some further evidence in support for our earlier conclusions on symmetry by explicitly examining for evidence of asymmetry in the growth cycle of the substitution effect. In Table 11 we present the results of testing for the presence of asymmetry. The results in Table 11 are based on calculating a measure of steepness and deepness of the growth cycle. Steepness (when contractions are steeper than expansions) and deepness (when troughs are deeper than peaks are tall) was first proposed by Sichel [1993].

Table 11

Tests for Deepness and Steepness

Ugs Ue cyc Ugs Ue cyc Deepness coefficient p-value Deepness coefficient p-value Steepness coefficient p-value Steepness coefficient p-value (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) Australia – 0.0036 0.641 0.00079 0.930 – 0.0011 0.892 0.0083 0.496 Belgium – 0.0046 0.757 – 0.01500 0.258 0.0048 0.555 0.0243 0.322 Canada – 0.0044 0.877 – 0.00560 0.608 – 0.0017 0.927 – 0.0056 0.608 Denmark – 0.0082 0.654 – 0.00440 0.709 – 0.0068 0.546 – 0.0071 0.409 Finland – 0.0089 0.363 0.00800 0.566 0.0063 0.530 0.0097 0.636 France – 0.0155 0.351 – 0.01090 0.362 – 0.0018 0.905 0.0002 0.985 Italy 0.0135 0.330 – 0.01230 0.286 0.0027 0.695 0.0150 0.432 New Zealand – 0.0017 0.864 – 0.00570 0.593 0.0043 0.600 0.0008 0.910 Norway 0.0020 0.817 – 0.01840 0.330 0.0005 0.944 0.0084 0.597 Portugal – 0.0179 0.469 – 0.00920 0.383 0.0039 0.664 – 0.0159 0.489 Sweden 0.0061 0.704 0.01180 0.514 0.0018 0.817 0.0083 0.343 Uk – 0.0050 0.555 0.00920 0.430 0.0078 0.234 0.0077 0.459

Tests for Deepness and Steepness

92The sample skewness statistics used to test for deepness examines the presence of negative skewness in the data series relative to it’s mean or trend. The coefficient of skewness is used as a test for steepness. From the result in Table 11, there appears to be no evidence of deepness or steepness in the growth cycle of the substitution effect component of the shadow economy. That is, the peaks in the cycles are found to be as tall as the troughs are deep and the contractions in the growth cycles are as steep as the expansions. These results would suggest that changes in the unemployment rate do in fact have very similar effects on changes in the substitution effect irrespective of whether they are positive or negative. The shadow economy, it seems, is a temporary refuge in periods of unemployment for those genuinely wanting to engage in legitimate employment. The increase in duration spells motivated by generous social security payments only postpones unemployment temporarily, at most by 3 quarters or 9 months. [33]

Conclusion

93The objective of this paper has been to investigate the relationship between the unemployment rate and the shadow economy. Previous literature on this topic has suggested that the relationship between these two variables is ambiguous, predominantly because there exists a heterogeneous group of people working in the shadow economy and there are also various cyclical forces at work such that they produce a net effect that is weakly correlated with unemployment. In this paper we have provided a suggestion for disentangling these cyclical effects, so as to study the component of the shadow economy that is influenced directly by those who are unemployed. This effect we referred to as the “substitution effect”, which typically increases during declining periods of legitimate economic activity (and increasing unemployment). Equipped with this approach for measuring the “substitution effect”, we discovered that a relationship exist between changes in the unemployment rate and shadow economy activity.

94By examining the growth cycle characteristics of the “substitution effect” component of the shadow economy we determined that the growth cycles are symmetric (in terms of steepness and deepness) and that changes in the unemployment rate, whether positive or negative, had similar impacts on changes in the substitution effect component. This suggests that the shadow economy is a source of financial support during periods of unemployment for those genuinely wanting to participate in the legitimate economy. Although this does not exclude the possibility that long-term unemployed may also be participating in the shadow economy, it would appear that short-term fluctuations in unemployment directly contribute to short-term fluctuations in the shadow economy.

95When we considered the various unemployment support programs across the 12 oecd countries, there appeared to be no real systematic relationship between the generosity of the social security systems and the nature of short-term shadow economy activity by the unemployed. Even the various replacement rates across the oecd countries appear to have little consequence on the rate at which the unemployed take-on and cut-back shadow economy activity. There is however some evidence to suggest that extended duration spell in unemployment lasts anywhere between less than 3 months to approximately 9 months.

96On the whole we could argue that dealing with unemployment participation in the shadow economy as a way of correcting the inequity it generates, is best handled by more stringent monitoring of those receiving unemployment benefits rather than reducing replacement rates as a way of encouraging re-integration into the work force. A strategy of reducing replacement rates would not only fail to maintain adequate support for those experiencing financial hardship during periods of unemployment, it is likely to have little impact on reducing participation by the unemployed who are willing and able to engage in shadow economy activity.