AIM-AP. Accurate Income Measurement for the Assessment of Public Policies. Citizens and Governance in a Knowledge-based Society

Transcription

1 Project no: AIM-AP Accurate Income Measurement for the Assessment of Public Policies Specific Targeted Research or Innovation Project Citizens and Governance in a Knowledge-based Society Deliverable 1.4a Home production and fringe benefits in Greece Due date of deliverable: January 2008 Actual submission date: March 2008 Start date of project: 1 February 2006 Duration: 3 years Lead partner: CERES Revision [draft]

2 Distributional effects of consumption of own production and fringe benefits: Greece Christos Koutsambelas Athens University of Economics and Business and CERES and 2 Panos Tsakloglou Athens University of Economics and Business, IZA and CERES PRELIMINARY VERSION: PLEASE DO NOT QUOTE WITHOUT THE AUTHORS PERMISSION 1 Department of International and European Economics Studies, Athens University of Economics and Business, Pattision Str 76, Derigny Wing,, 5 th floor, Athens 10484, Greece. ckouts@aueb.gr 2 Corresponding author: Department of International and European Economics Studies, Athens University of Economics and Business, Pattision Str 76, Athens 10484, Greece. E- mail:tsaklog@aueb.gr 1

3 Table of Contents 1.Introduction Consumption of own production and fringe benefits in Greek economy Dataset and Methodology Empirical results: Reassessing inequality Distributional statistics Changes in the level of inequality Graphical Approach: Lorenz and Concentration Curves Inequality decomposition by subgroups Inequality decomposition by sources Changes in the level of poverty Poverty decomposition by subgroups Conclusion References Tables

4 1.Introduction As it is well regarded in economic science, individuals derive welfare from full income, which is nothing else from the sum of monetary and non-monetary income. Despite that monetary income is a conceptually simple notion, this is not the case for non-monetary income. That is why the classic Haig-Simons definition of income (Simons, 1938), which states that income is the sum of the market value of property rights exercised in consumption and the changes in the value of the store of such rights from the beginning until the end of the period the income is measured, is not trivial at all. Trying to fully satisfy this definition in empirical works is not easy and definitely depends upon the informational completeness of the used dataset. In other words the problems that are inherent in the measurement, valuation and imputation of non pecuniary incomes make their omission sometimes a necessity. Unfortunately it is not a harmless omission, since the distribution of in-kind goods may vary systematically across by country, by location or by population subgroup. Consequently estimates reported in empirical income studies may be seriously biased. Moreover, this practice has important implications for the design of policies aiming to fight poverty and social exclusion, since it may lead to imperfect targeting and misallocation of resources. This is the reason that in recent years many leading researchers and international organizations have called for the development of methods providing reliable estimates of non-cash income components and their incorporation in distributional studies [Hagenaars et al (1994), Atkinson et al (1995), van de Walle and Nead (1995), Canberra Group (2001), Atkinson et al (2002)]. In same spirit many income studies such as Smeeding et al (1993), Evandrou et al (1993), Harding et al (2006), Garfinkel et al (2006) and many other similar works, try to implement a more comprehensive definition of income by incorporating in their estimates as many sources of non-cash income as possible. There are also studies that choose to focus on the changes that are induced in income distribution by the inclusion of a specific non-monetary income such as education; Tsakloglou and 3

5 Antoninis (1999), imputed rents; Frick and Grabka (2003), health; LeGrand (1978) etc. This study belongs to the latter group of the literature, for its intention is to estimate the distributional impact of consumption of own farm and non-farm production and fringe benefits on the income distribution, also in the Appendix can be found estimates of the distributional impact of non-monetary transfers received from other households. What makes this study from distributional point of view interesting is the implicit evidence that their distribution effectively differs from the distribution of monetary income. Intuitively the inclusion of the imputed value of household production and fringe benefits should affect both tails of the distribution. Consumption of home production is more prevalent to low-income households (especially in the case of own farm production, see for example Bryant and Zick (1985)), in contrast with fringe benefits that are concentrated disproportionately to high wage employees (For example Pierce (2001) reports that fringe benefits are less equally distributed than wages). Therefore the inclusion of these non-money incomes in an empirical income study could not only but enrich the analysis, by shedding some more light in the measurement of poverty and inequality in Greece. To do so we take advantage of the rich informational basis of the 2004 Greek Budget Household Survey (hereafter BHS) in order to estimate the distributional effects of consumption of own production and fringe benefits in Greece. The structure of the paper goes as follows; Section 2 discusses the relevance of other non-cash incomes (hereafter under this generic term we will describe income advantages derived from own farm and non-farm production and fringe benefits) in Greek economy emphasizing issues such as the socioeconomic origin of their receivers and their composition with respect of each consumption category, section 3 briefly describes our dataset and methodological settings, section 4 reports estimations of poverty and inequality before and after the inclusion of other non-cash incomes and other distributional statistics as well as other distributional statistics and finally section 5 concludes. 4

6 2.Consumption of own production and fringe benefits in Greek economy. Graph 1 reports the position of the receivers of imputed incomes in the income distribution when the population is grouped in quintiles according to their equivalent disposable income. An immediate point is that the distribution of the receivers of imputed income derived from own farm production is clearly skewed towards the bottom of the distribution (about one quarter of the receivers belongs to the poorest quintile), exactly the opposite pattern is observed for fringe benefits. Fringe benefit receivers are clearly concentrated towards the upper part of the distribution. Finally a rather erratic pattern characterizes the distribution of the receivers of imputed income from own non-farm production. Finally graph 1 also describes the distribution of all receivers, irrespectively of the non-cash income enjoyed, and it more or less reflects the points made earlier; most receivers are concentrated in the poorest quintile (a result caused by own farm production), however many receivers can be found also in the richest quintile (mostly fringe benefits receivers). Of course these estimates try to capture only the location of receivers and naturally neglect the relative size of imputed income to their disposable income. Moreover the estimates may be driven from demographics. Therefore any distributional judgement at this stage would be premature. Nevertheless intuitively we can expect that fringe benefits are pro rich, own farm production pro poor and own non-farm production follows a middle path, mildly inclined towards progressivity. Table 1 disaggregates the population using the following socioeconomic criteria: household type, occupational status of the household head, educational status of the household head and geographical area. The first column reports the corresponding population shares of each subgroup, the next three contain estimates of the proportion of population of the specific subgroup that enjoy a positive imputed income from own farm production, own non-farm production and fringe benefits respectively. Consumption of own farm production is very common for people living in rural areas 5

7 (about 73% of the people that live in rural areas enjoy imputed income from own farm production), percentages are also high for people that belong to households headed by a low educated (44%), pensioner (39%), self employed (42%). In contrast consumption of own non-farm production percentages is less common, except of the case of the self-employed group, whereas a significant proportion of the group receives them. The latter also reflect an important aspect of the structure of Greek economy; a considerable portion of Greek labor force is not only employed in the sectors of agriculture, cattle breeding and fishing [11.4% of total labor force, BHS (2004)], but also most of them are either self-employed (71% ) or work in their family enterprise (22%). Consequently the widespread non-monetization of the production of the farm sector is not surprising. Finally fringe benefits, as it was expected, are more prevalent for the households headed by white (15.3%) and blue collar workers (14.9%), high educated (11.74% for the tertiary education group and 10.64% for the upper secondary education one) and for people located in urban areas (9.71%). Table 2, reports shares of households with non-purchased consumption above 10 euros per consumption category. Consumption of own farm production mainly consists of food and in some cases of housing goods (for example firewoods), consumption of own non-farm production (that is consumption of goods produced from household's own entreprise) naturally includes a greater variety of goods such as food, clothing, durables etc. Also fringe benefits comprise of several types of goods. But in most cases they take the form of the free provision of hotel/restaurant or communication services, even if there is a significant number of food provision. Finally table 3 that reports shares of non-purchased consumption in total consumption per category shows that consumption of non-purchased food represents almost 10% of the total food consumption, making non-purchased food - irrespectively of its source - as the most important case of non-purchased consumption. Less important, at least with regard their impact on household budget, are non-purchased communication services (a common fringe benefits as it was revealed by table 1) and alcohol/tobacco (2.8% and 2.0% of total consumption respectively). Finally the shares of the 6

8 remaining non-purchased consumption categories are too small to worth commenting them. 3.Dataset and Methodology The data used in the paper are the micro-data of the 2004/5 Greek Household Budget Survey, which was carried out by the National Statistical Service of Greece. The survey covers all the private (non-institutional) households of the country and its sampling fraction is 2/1000 (around 6,500 households or 18,000 individuals). It contains detailed information about consumption expenditures (actual and imputed), incomes and socioeconomic characteristics of the households and the household members. The baseline distribution is the distribution of disposable income. All monetary values were expressed in constant mid-2004 values in order to remove the impact of inflation. The distributions used are distributions of equivalent household disposable income per capita and they are derived using the modified OECD equivalence scales (Hagenaars et al, 1994) that assign weights of 1.00 to the household head, 0.50 to each of the remaining adults in the household and 0.30 to each child (person aged below 14) in the household. Since the estimates in the HBS are expressed in monthly figures, the cost estimates of Table 1 are adjusted accordingly. The estimation of the value of own production and fringe benefits depends ultimately upon the information provided by the dataset. Fortunately the Greek budget survey provides the necessary material for a rich analysis. Imputed own production and fringe benefits are included both in consumption and income information of the household. According to the classification of the survey, there are five ways to obtain a commodity: (1) Purchase (2) From own farm production 7

9 (3) From own non-farm production (services not included) (4) From other households (for example charities or gifts) (5) From employer Also the income list of the survey contains the follow five imputed items (excluding imputed rents) (1) Imputed income derived from the production of own agricultural or cattlebreeding production. (2) Imputed income derived from goods that are produced by an enterprise that belongs to the household. (3) Imputed value of goods that were given for free to the household. (4) Imputed income from fringe benefits except of cars. (5) Imputed value of car services provided by the employer. Imputed income (3) mainly refers to gifts and charities and was not included in the subsequent analysis. Incomes (4) and (5) were merged, for there were very few cases of imputed income (5) (45 observation). Consequently there are three categories of imputed income to consider, the first two refer to own production and the third to fringe benefits. Moreover the distributional impact of item (3) which was excluded from the main body of the analysis, will be incorporated in the tables provided but not commented. 4. Empirical results: Reassessing inequality 4.1 Distributional statistics In this subsection we report estimates of mean transfers per quintile, proportional increases in equivalent disposable income and changes in income shares induced by the inclusion of imputed incomes. In a sense these distributional statistics are precursors of inequality and poverty results that will follow in the next sections. Table 4 examines the mean imputed income per capita for each quintile. One can easily 8

10 notice that the mean income from own farm production not only is of considerable size, but also is higher in poor quintiles than in the rich one. Average imputed income stemming from own non-farm production follows a quite opposite pattern, first of all its size is modest and secondly is slightly higher for the top quintiles. Unsurprisingly average fringe benefits are high for the household members that belong to the rich quintiles and negligible for the poor ones. When all items taken together, the average aggregate imputed income is up to 13 euro. This number doesn't differ significantly across quintiles, indicating that somehow the different categories of imputed income complement each other, for example low-income households enjoy low fringe benefits but rather high consumption from own production and vice versa for the high-income households. As a result the level of average imputed income doesn't fluctuate much across quintiles. However from distributional point of view, what matters are the changes of the relative positions of individuals, something that is better captured from Table 5, which provides estimates of the proportional rise in the income of the quintiles caused by the inclusion of the imputed income. In both imputed income from own farm production and own non-farm production, the increase in the income of the quintiles diminishes as we move up the income distribution. Thus imputed income from own farm production causes an 12% increase in the equivalent disposable income of the poorest quintile, this increase declines steadily ass we move to the top quintiles, reaching 4.9% in the case of the top quintile. Imputed income from own non-farm production causes more significant increases in both tails of the distribution than in the middle, reflecting merely the fact that middle income classes in Greece are comprised mainly of blue and white collar workers (Note also from Table 1 that in these groups are income from own non-farm production is not very common). Naturally such patterns is not observed in the case of fringe benefit. Finally if we aggregate all imputed incomes, they account for 7% of the income of the bottom quintile, the corresponding share declining gradually as we proceed to higher quintiles, reaching 1.0% in the case of the top one. An immediate conclusion is that this declining pattern of proportional increases indicates that other non-cash income are progressively distributed across income distribution. 9

11 Table 6 reports the quintile income shares before and after the inclusion of imputed incomes in the concept of resources that are available to the household members. After their inclusion, the shares of the three lowest quintiles, and especially the bottom, increase (from 7.5% to 7.8%, 12.7% to 12.9% and 17.0% to 17.1% for each of the three bottom quintiles) while that of the top two decline (from 22.94% to 22.9% for the fourth quintile and from 39.8% to 39.4% for the top one). Again the main driving force behind these changes seems to be the imputed income derived from own farm production. 4.2 Changes in the level of inequality Table 7 examines the impact of other non-cash income on aggregate inequality; that is, it reports the proportional change in a number of inequality indices when we move from the distribution of disposable income to the distribution of augmented income. As inequality indices we chose the widely used Gini index and two members of the parametric family of Atkinson (1970) indices. The value of the inequality aversion parameter in the latter is set at (e=0.5 and e=1.5). Both indices satisfy the desirable properties for an inequality index (anonymity, mean independence, population independence, transfer sensitivity). Higher values of e make the Atkinson index relatively more sensitive to changes closer to the bottom of the distribution while, in practice, the Gini index is relatively more sensitive to changes around the median of the distribution [Cowell (2000), Lambert, (2001)]. When moving from the distribution of disposable income to the augmented distribution of resources, the Gini index declines by 2.0%, while the two Atkinson indices (e=0.5 and e=1.5) decline by around 4.1% and 4.9% respectively. Almost the entire effect is driven by the progressive redistributive impact of the consumption of own farm production ( for example the Gini index is reduced by about -1.8% and the two Atkinson -3.7% and -4.3% when we consider only for the partial effect of imputed income from own farm production). The other two sources of imputed income that is consumption of own non-farm production and fringe benefits have only marginally impacts on inequality. 10

12 4.3 Graphical Approach: Lorenz and Concentration Curves A well-known way of summarizing and comparing income distribution is the Lorenz curves. Lorenz curve plots cumulative proportions of income units (ordered according to their income) against cumulative proportions of income received by these income units. Algebraically it is defined as: L(p)= xf(x)/μ dx, p Ε [0,1] where x is the income variable, p is the proportion of population and f(x) the frequency density function. For the purpose of our analysis we define a function d(p), which is the difference between the Lorenz curve of the income distribution when each of the imputed income categories is included and the Lorenz curve of the baseline distribution (i.e. the Lorenz curve for the equivalent disposable income) that is: di( p) = Li ( p) Lbase( p). Thus, Graph 2 depicts: d 1(p), when L1(p) is the Lorenz curve for the distribution of disposable income plus the imputed income derived from of own farm production. d 2(p), When L2(p) is the Lorenz curve for the distribution of disposable income plus the imputed income derived from household's enterprise. d 3(p), When L3(p) is the Lorenz curve for the distribution of disposable income plus the fringe benefits. d all (p), When L 3 (p) is the Lorenz curve for the distribution of disposable income plus all imputed incomes. If the d-function is above zero for every possible value of p, then we have a case of Lorenz dominance that allows us to safely infer that inequality is reduced according to every possible index of relative inequality [Atkinson, (1970)] and vice versa. Graph 2 is very illuminating of the distributional implications of the inclusion of other non-cash incomes in the baseline distribution. An immediate conclusion is that the 11

13 augmented distribution Lorenz dominates the baseline distribution ( d all (p) is positive everywhere in its domain). This impact is almost entirely driven by the inclusion of imputed income from own farm production, something reflected on the similarity of d all (p) and d 1(p) curves. In fact dall(p) has almost exactly the same shape as d 1 (p), but is located a bit higher, for its redistributive impact is mildly enhanced by the equalizing effect of imputed income derived from own non-farm production. As far as d 2 (p) is concerned, it is everywhere positive (once again a case of Lorenz dominance). However d 3 (p) follows a different pattern; it crosses once the horizontal axis, being negative for p<0.79 and positive for p>0.79. This Lorenz crossing suggest that we cannot infer any unambiguous conclusion regarding the distributional impact of fringe benefit. Whatever conclusion will be conditional on distributional tastes. Given that we have more inequality at the bottom and at the middle and less at the top of the distribution after the inclusion of fringe benefits, an inequality index which gives more weight to transfers to the upper tail of distributions (a possible such index would be whatever belongs to the GE family if we set a high enough value of the parameter) could yield an inequality reduction (and vice versa). However it should be admitted that in order to measure a decrease in inequality, we have to set an unreasonably high value of the parameter (technically this happens because the crossing point is also high). The above results are also verified by the use of concentration curves analysis. The logic of concentration curves is similar to that of the Lorenz curves with the crucial difference that we now plot shares of one variable against quantiles of some other variable. In this sense, Lorenz curves are just a subset of the set of all possible concentration curves (for example, the Lorenz curve for disposable income is nothing else than the concentration curve for disposable income with respect to disposable income). The concentration curves are useful in that they allow us to isolate graphically and compare the pattern of distribution of each of the non-cash components in question. The corresponding concentration curves are reported in Chart 3. 12

14 C (p), when L 1 1(p) is the Lorenz curve for the distribution of disposable income plus the imputed income of own farm production. C (p), when L 2 2(p) is the Lorenz curve for the distribution of disposable income plus the imputed income of own non-farm production. C (p), when L 3 3(p) is the Lorenz curve for the distribution of disposable income plus the fringe benefits. C all (p), when L 3 (p) is the Lorenz curve for the distribution of disposable income plus all imputed incomes. Examining Graph 3, we notice that C 1 (p) lies well above the 45 o, this means that imputed income derived from own farm production is disproportionately distributed in favour of the people that belong to the low part of the distribution. C 2 (p) is located near the 45 o, that is imputed income derived own non-farm production is distributed evenly across distribution and finally C 3 (p) is bellow the Lorenz curve for the baseline distribution for p<0.69 and above it for p>0.69, thus for the bottom part of the income distribution fringe benefits are even more unequal than the distribution of disposable income but not so for the upper part. Finally C all (p), which is the combined effect of the above concentration curves, lies above the Lorenz curve for the baseline distribution. This is also one of the main points of this study; the distribution of other non-cash incomes is more equally distributed than the distribution of disposable income. 4.4 Inequality decomposition by subgroups The above results mainly refer to aggregate effects in inequality induced by other non-cash incomes. But what is also of potential interest is the source and structure of inequality and how they change by the inclusion of imputed incomes in the definition of income. An insight in these aspects of inequality can be obtained by exploiting the property of decomposability that indices of the Generalized Entropy family possess. Therefore, we performed decomposition by population subgroups on the mean log 13

15 deviation index. The choice of subgroups has been made with respect to their relevance in the analysis and what is available in our dataset, so we decomposed by household type (older single persons or couples, younger single persons or couples, couple with children up to 18, mono-parental household, other household type), by socioeconomic status of the household head (blue collar worker, white collar worker, self-employed, unemployed, pensioner, other), by education level of the household head (tertiary education, upper secondary education, lower secondary education, primary or less) and by age (bellow 25, 25-64, over 64). Table 8 reports the results. The first column of the table shows the different partitions of population we adopted and the various subgroups that came up. Column A reports the population share of each subgroup, so that we have an idea of its relevant importance; columns B and C contain the estimates of MLD for the baseline distribution (distribution of equivalent disposable income) and the one which includes imputed incomes, column D reports the percentage change of the inequality index. Columns E, F, G, H report the contributions to inequality and the relative positions of each subgroup before and after the inclusion of imputed incomes respectively and finally the last column refers to changes in relative position. The structure of inequality does not change substantially as we move from the distribution of disposable income to the augmented income distribution. Inequality declines within all population sub-groups and in all groupings both within and between inequality also declines. Finally with regard relative positions, we only observe marginal changes. Only members of households headed by a self-employed or a low educated person enjoy a proportional change of relative position higher than 1% (1.3% and 1.6% respectively), all the other changes are marginally. 4.5 Inequality decomposition by sources In this section we disaggregate the full income of individuals (that is the equivalent income of individuals after the inclusion of non-cash income) in the following factor 14

16 components; disposable income, imputed income from own farm production, imputed income from household's enterprise and fringe benefits in order to assess the contribution of these sources to total inequality. Following Pyatt et all (1980) Gini index can be written as: written as: K mk G = Σ Rk m G k = 1 k (1) m m where k R and are the mean of component k and total income respectively. k is the relative correlation coefficient of component k (that is the ratio of the covariance of component k with the total income rank and the covariance of component k with its G own rank) and finally k is the Gini index for the component k. (1) can be rearranged (divide both parts of the equation by G) so as to yield: K Σ w 1 kg = k (2) k= 1 w where k g is the share of component k in total income and k is the relative concentration coefficient of component k in aggregate inequality. The product wg k k can be interpreted as the proportional contribution of component k in aggregate g inequality. From the estimation of k we can detect whether an increase in the g income component k will increase or decrease aggregate inequality. In fact if k > 1 g then inequality will increase and if k <1 inequality will decrease. Furthermore from (1) we can calculate the elasticity of inequality with respect to a proportional change in component k. e dg m = k k wkg dmk G = k w k w g Estimates of k, k e, k are reported in Table 9 for each component and for all components taken together and for various values of the parameter of distributional g sensitivity ν (ν=1.5, ν=2, ν=4). The estimates of k show that all components mitigate aggregate inequality whatever the value of the distributional sensitivity parameter, 15 (3)

17 gs since all k g are less than one, except of the fringe benefits whose k is more than one whatever the value of the parameter v. These estimates taken together with the values are consistent with the estimations of elasticity of inequality with respect to components k. Indeed elasticity is always negative except of the case of fringe benefits. Elasticity of inequality of aggregate other non-cash incomes is negative (naturally fringe benefits are not so high so as to undo the equalizing effect of imputed income from own farm production and household's own entreprise. Thus increasing marginally the size of the imputed income, ceteris paribus, will decrease aggregate inequality, a decrease that is more significant the higher the value of the parameter v. wk 4.6 Changes in the level of poverty Poverty rates in Greece are considered high, especially in comparison with EU countries. In this section, we apply the parametric family of FGT(a) index in order to find out whether the inclusion of other non cash income does reduce poverty. The α parameter is defined as the social aversion to poverty; the higher the value of alpha the higher the aversion to poverty. When the value of the poverty aversion parameter is set at α=0, the index becomes the widely used poverty rate, that is the share of the population falling below the poverty line. When α=1, the index becomes the normalized income gap ratio, while when α=2 the index satisfies the axioms proposed by Sen (1976) (anonymity, focus, monotonicity and transfer sensitivity) and is sensitive not only to the population share of the poor and their average poverty gap, but also to the inequality in the distribution of resources among the poor. Following the practice of Eurostat, the poverty line is set at 60% of the median income. Table 10 reports the corresponding results. The results reported in Table 10 suggest that poverty is reduced according to all versions of FGT index after the inclusion of imputed income in the concept of resources. The poverty reducing effect of other non cash-income is enhanced for 16

18 higher values of the poverty sensitivity parameter alpha; when α=2 recorded poverty is reduced by almost 20%. Once again, the bulk of the poverty reduction can be attributed to the imputed income derived from own farm production, while the poverty reducing effect of imputed income from own non-farm production is mild and the partial effect of fringe benefits even causes a marginal increase in poverty for values of alpha higher than zero. 4.7 Poverty decomposition by subgroups This section reports the results of the poverty decomposition by subgroups, by which we try to identify population groups at high of poverty risk and how this risk and their contribution to aggregate poverty after the change in the concept of resources. The task is accomplished thanks to the property of additive decomposability that FGT index possess and the results are reported in Table 11. The differentiation of population was done by the same criteria as in the preceding inequality decomposition analysis. Column A reports population shares for each group, columns B and C contain estimates of FGT (a=0), D percentage changes in poverty, E and F the contributions of each subgroup to aggregate poverty, G, H, I, J, K report the same magnitudes but for FGT (a=1) and finally L, M, N, O, P for FGT (a=1). In general, after the inclusion of other non-cash income in the concept of resources the poverty risk declines in almost all population sub-groups; apart of the people that belong to households headed by unemployed or white collar workers. the tenants. The poverty reductions are especially high for the households headed by self-employed or pensioners and the low educated (lower secondary education and primary education or less groups). For example for the people that belong to households whose head is self employed, poverty falls -11.6%, % and % for the FGT a=0,1,2 respectively, for the primary education or less group the corresponding figures are - 7.8%, % and % and finally for the lower secondary education group %, -7.46%, %. The significant poverty reductions observed for these highly overlapping groups can be merely attributed to the fact that self employment is widespread among Greek farmers. 17

19 6.Conclusions The main aim of this paper was to assess the impact of other non-cash income, namely income derived from own farm production, own non-farm production and fringe benefits on income distribution. The empirical findings one more time in the literature of empirical income studies demonstrate the importance of including noncash incomes in applied welfare analysis. For the distribution of these resources may systematically differ from the distribution of monetary resources, thus making their omission a source of potential bias in the estimation of measures of relative well being This is the case with non-monetized production and non-pecuniary employers compensation in Greece. We found that the distribution of own production is more equally distributed than disposable income, in contrast with fringe benefits that appear to be concentrated more on the upper part of the distribution. This result is especially strong in the case of own farm production, something which can be attributed to the widespread self employment of Greek farmers. Consequently both aggregate inequality and poverty decrease after the inclusion of other non cash imputed incomes. Finally using decomposition analysis we concluded that inclusion of other non cash incomes is, at least in relative terms, more beneficial for some socioeconomic groups of the population that are found to be more sensitive to poverty (farmers, people living in rural areas, elderly, Tsakloglou (2000), we are inclined to support that at least in Greece non-monetization can be considered as a mean of self defence against poverty and inequality. 18

20 7. References Atkinson A.B. (1970). "On the measurement of inequality", Journal of Economic Theory 2, pp Atkinson, A.B., Cantillon, B., Marlier, E. and Nolan, B. (2002) Social indicators: The EU and social inclusion, Oxford University Press, Oxford. Atkinson, A.B., Rainwater, M. and Smeeding, T. (1995), Income distribution in OECD countries, OECD, Paris. Bryant, W. K. and Zick C. D., (1985), Income distribution implications of rural household production, Journal of Agricultural Economics, Vol. 67, No. 5, Proceedings issue., pp Canberra Group (2001), Final Report and Recommendations, The Canberra Group: Expert Group on Household Income Statistics, Ottawa. Cowell F.A. (2000). Measurement of inequality in A.B. Atkinson and F. Bourguignon Handbook of Income Inequality, Vol. I, North Holland, Amsterdam. Evandrou M., J. Falkingham, J. Hills and J. Le Grand (1993), Welfare benefits in kind and income distribution, Fiscal Studies, 14. Frick, R. and Grabka, M., (2003) Imputed Rent and Income Inequality: A Decomposition Analysis for Great Britain, West Germany and U.S., The Review of Income and Wealth, 49 (4), Garfinkel I., Rainwater L., and Smeeding T.M. (2006) A Reexamination of Welfare State and Inequality in Rich Nations: How In-Kind Transfers and Indirect Taxes Change the Story Journal of Policy Analysis and Management 25, pp Hagenaars, A.J.M., de Vos, K. and Zaidi, M.A. (1994). Poverty statistics in the late 1980s: Research based on micro-data. Luxembourg: Eurostat. Harding A., Lloyd R. and Warren N. (2006). "Moving beyond traditional cash measures of economic well-being: including indirect benefits and indirect taxes", National Centre For Social and Economic Modelling, Discussion Paper no. 61, University of Canberra. Lambert P.J. (2001). The distribution and redistribution of income: A mathematical analysis, 3rd edition, Manchester University Press, Manchester. LeGrand J. (1978) The distribution of public expenditure: The case of Health Care, Economica, New Series, Vol.45, No.178. (May 1978), pp National Statistic Service of Greece. Household Expenditure Survey 2004/2005. Athens: NSSG 19

21 Pierce B., (2001). Compensation Inequality, Quarterly Journal of Economics, vol 116., pp Pyatt, G., Chen, C., and Fei, J. (1980). The distribution of income by factor components., Quarterly Journal of Economics, vol 95., pp Sen A.K. (1976). "Poverty: An ordinal approach to measurement", Econometrica 44, pp Simons, Herbert (1938). Personal Income Taxation, Chicago: University of Chicago Press Smeeding T.M., Saunders P., Coder J., Jenkins S.P., Fritzell J. Hagenaars A.J.M. Hauser R. and Wolfson M. (1993). "Poverty, inequality and living standard impacts across seven nations: the effects of non-cash subsidies for health, education and housing", Review of Income and Wealth 39, pp Tsakloglou P. and Antoninis M. (1999) On the distributional impact of public education: evidence from Greece, Economics of Education Review 18, pp Tsakloglou, Panos Poverty and Anti-Poverty Policies in Greece and a Comparison with Other Mediterranean EU Member States. In Contemporary Greece and Europe, edited by A. Mitsos, and E. Mossialos. Ashgate, Aldershot, pp Van de Walle, D. and Nead, K. (eds) (1995) Public spending and the poor: Theory and evidence, Johns Hopkins University Press, Baltimore and London. 20

22 8. Charts and Tables Chart 1: Distribution of imputed income receivers per quintile 0,35 0,30 0,25 0,20 0,15 0,10 0,05 Imputed income from ow n farm production Imputed income from ow n non-farm production fringe benefits All imputed incomes 0, Quintiles 21

23 Table 1: Shares of people who receive positive other non-cash incomes per 3 socioeconomic category of whom positive imputed income from Characteristic of household or household head Household type Older single persons or couples (at least one 65+) Younger single persons or couples (none 65+) Couple with children up to 18 (no other HH members) Mono-parental household Population Share own farm production own non-farm production fringe benefits 7,86 39,01 1,44 0,06 18,02 26,02 2,29 4,74 33,81 23,13 6,58 10,10 1,52 11,15 2,33 5,93 Other household types 39,31 35,96 7,70 9,64 Socioeconomic group of HH head Blue collar worker 23,43 19,13 2,76 15,31 White collar worker 15,00 19,15 2,06 14,90 Self-employed 23,32 41,64 15,02 3,50 Unemployed 2,31 16,21 6,71 1,92 Pensioner 28,01 38,84 3,74 2,98 Other 8,44 18,50 1,80 7,56 Educational level of HH head Tertiary education 20,50 15,30 3,47 11,74 Upper secondary education 27,20 21,01 7,63 10,64 Lower secondary education 13,05 28,98 4,79 8,48 Primary education or less 39,76 43,38 6,04 4,40 Location urban 66,72 14,60 4,52 9,71 semi urban 12,81 4,08 8,20 7,43 rural 20,47 72,88 8,53 3,54 ALL ,88 5,81 8,16 3 Only positive imputed incomes over 10 euros are considered. 22

24 Table 2. Share of households with positive non-purchased consumption per category 4 from own farm from own non- farm from other households from employer Consumption group production production Food 26,75 2,26 16,60 1,39 40,64 Alcohol & Tobacco 3,29 0,50 1,20 0,09 5,00 Clothing & Footwear 0,00 0,39 0,46 0,27 1,12 Housing etc 3,43 0,56 2,65 0,10 6,71 Durables 0,00 0,59 0,63 0,07 1,29 Health & Personal Care 0,00 0,00 0,19 0,00 0,19 Transport 0,00 0,06 0,76 0,28 1,09 Communications 0,00 0,00 0,96 2,15 3,07 Recreation & Culture 0,00 0,09 0,59 0,07 0,73 Education 0,00 0,00 0,42 0,03 0,45 Hotels, Restaurants etc 0,00 1,37 8,54 2,43 12,10 Other Goods & Services 0,00 0,65 1,59 0,31 2,49 All Table 3. Share of non purchased households with positive non-purchased consumption in total consumption per category 5 Consumption category from own farm production from own non- farm production from other households from employer All other Food 5,5 0,7 2,9 0,1 9,3 Alcohol & Tobacco 1,6 0,5 0,6 0,1 2,8 Clothing & Footwear 0,0 0,5 0,2 0,3 1,0 Housing etc 0,3 0,0 0,6 0,0 0,9 Durables 0,0 0,2 0,7 0,0 1,0 Health & Personal Care 0,0 0,0 0,2 0,0 0,2 Transport 0,0 0,1 0,3 0,1 0,4 Communications 0,0 0,0 0,4 1,6 2,0 Recreation & Culture 0,0 0,0 0,3 0,1 0,4 Education 0,0 0,0 1,1 0,1 1,2 Hotels, Restaurants etc 0,0 1,5 5,1 1,7 0,6 Other Goods & Services 0,0 0,1 0,3 0,2 0,5 4 5 Only positive imputed incomes over 10 euros are considered. Only positive imputed incomes over 10 euros are considered. 23

25 Table 4: Equivalent mean imputed income per capita per quintile Quintiles from own farm from own non- farm from other households from employer All except of from other production production households 1 12,0 2,6 14,4 0,7 15,3 29,7 2 9,6 2,4 11,0 1,3 13,4 24,4 3 6,6 1,9 7,5 2,2 10,7 18,2 4 6,1 3,2 6,7 3,9 13,2 19,9 5 4,9 2,9 8,5 4,4 12,2 20,7 All 7,9 2,6 9,6 2,5 13,0 22,6 All Quintiles Table 5: Proportional increase in disposable income per quintile due to imputed incomes from own farm production from own non- farm production from other households from employer All except of from other households 1 5,2 1,3 5,5 0,3 6,8 12,3 2 2,5 0,7 2,5 0,4 3,6 6,1 3 1,3 0,4 1,4 0,4 2,1 3,5 4 0,9 0,5 0,9 0,6 2,0 2,9 5 0,4 0,2 0,7 0,4 1,0 1,7 All 1,3 0,5 1,4 0,4 2,2 3,6 All 24

26 Table 6. Quintile income shares before and after the inclusion of imputed incomes from own from own All except of from other from Quintiles Baseline farm non- farm from other All households employer production production households 1 7,5 7,73 7,53 7,67 7,48 7,8 7, ,69 12,84 12,73 12,81 12,68 12,9 12, ,07 17,12 17,03 17,11 17,06 17,1 17, ,94 22,84 22,95 22,88 22,97 22,9 22, ,81 39,48 39,76 39,53 39,82 39,4 39,14 Chart 2. Differences between Lorenz curves 25

27 Chart 3. Concentration curves 26

28 Table 7. Proportional changes in inequality indices due to imputed incomes Inequality Iindex Baseline from own farm production from own non- farm production from other households from employer All except of from other households Gini 0,3217-1,8-0,3-1,5 0,1-2,0-3,4 Atkinson0.5 0,0849-3,7-0,6-3,1 0,1-4,1-7,1 Atkinson1.5 0,2406-4,3-0,9-3,9 0,2-4,9-8,5 All 27

29 Table 8. Inequality Decomposition by Population Subgroups Characteristic of household or household head A Β C D E F G H I Household type Older single persons or couples (at least one 65+) 7,8 0,1461 0,1379-5,60 6,4 6,3 71,4 71,8 0,6 Younger single persons or couples (none 65+) 18 0,2241 0,2173-3,06 22,7 22,9 98,8 98,8 0,0 Couple with children up to 18 (no other HH members) 33,6 0,1778 0,1711-3,78 33,6 33,6 103,4 103,0-0,4 Mono-parental household 1,5 0,1980 0,1883-4,88 1,7 1,7 81,7 81,1-0,8 Other household types 39,1 0,1506 0,1441-4,26 33,1 33,0 104,1 104,4 0,3 Within groups inequality 0,1734 0,1667-3,89 1,0 1,0 Between groups inequality 0,0044 0,0043-2,25 0,0 0,0 Socioeconomic group of HH head Blue collar worker 23,3 0,0920 0,0916-0,42 12,1 12,5 88,3 87,9-0,5 White collar worker 14,9 0,1051 0,1041-1,01 8,8 9,1 137,1 135,7-1,0 Self-employed 23,3 0,2652 0,2463-7,12 34,8 33,6 108,2 109,6 1,3 Unemployed 2,3 0,1342 0,1310-2,37 1,7 1,8 71,0 70,5-0,7 Pensioner 27,9 0,1667 0,1593-4,41 26,2 26,0 89,5 89,7 0,2 Other 8,4 0,1638 0,1607-1,90 7,7 7,9 86,6 86,0-0,7 Within groups inequality 0,1628 0,1561-4,09 0,9 0,9 Between groups inequality 0,0150 0,0148-1,26 0,1 0,1 Educational level of HH head Tertiary education 20,4 0,1384 0,1373-0,81 15,9 0,1 147,1 145,2-1,3 Upper secondary education 27 0,1410 0,1372-2,72 21,4 0,1 101,0 100,8-0,2 Lower secondary education 13 0,1444 0,1393-3,50 10,6 0,1 89,2 89,0-0,2 Primary education or less 39,5 0,1627 0,1551-4,70 36,2 0,1 78,5 79,8 1,6 Within groups inequality 0,1474 0,1430-2,96 0,8 0,8 Between groups inequality 0,0304 0,0279-8,20 0,2 0,2 Age of population member Below ,1697 0,1630-3,99 25,8 25,7 95,7 95,6-0, ,5 0,1714 0,1651-3,67 50,6 50,7 109,2 109,1 0,0 Over 64 20,6 0,1766 0,1684-4,64 20,5 20,3 82,2 82,4 0,3 Within groups inequality 0,1719 0,1651-3,94 1,0 1,0 Between groups inequality 0,0059 0,0058-1,35 0,0 0,0 A: Population Share B: Mean Log Deviation (Disposable Income) C: Mean Log Deviation (Disposable Income + Imputed incomes) D: % Change in Inequality E: % Contribution to Aggregate Income Inequality (Disposable Income) F: % Contribution to Aggregate Income Inequality (Disposable Income + Imputed incomes) G: Relative Income Position (Disposable Income) H: Relative Income Position (Disposable Income + Imputed Income) I: % Change in Relative Position 28

30 Table 9. Inequality Decomposition by factor components ν=1,5 ν=2 v=4 wk gk ek wk gk ek wk gk ek Baseline 0,9649 1,03 0,0266 0,9649 1,03 0,0258 0,9649 1,01 0,0129 from own farm production 0,0125-0,27-0,0160 0,0125-0,29-0,0162 0,0125-0,27-0,0159 from own nonfarm production 0,0045 0,69-0,0014 0,0045 0,72-0,0013 0,0045 0,79-0,0009 from other households 0,0140 0,24-0,0106 0,0140 0,28-0,0101 0,0140 0,36-0,0089 from employer 0,0042 1,33 0,0014 0,0042 1,42 0,0017 0,0042 1,44 0,0018 All 0,0351 0,24-0,0266 0,0351 0,26-0,0258 0,0351 0,32-0,0239 Table 10. Proportional changes in poverty indices due to imputed incomes Poverty Index Baseline from own from own All except of from other from farm non- farm from other households employer production production households All FGT0 0,3217-6,1-1,1-3,4-0,5-7,1-10,2 FGT1 0,0849-8,2-1,9-5,9 0,3-9,3-14,2 FGT2 0, ,0-2,9-8,6 0,4-12,9-19,9 29

31 Table 10. FGT Decomposition by Population Subgroups Characteristic of household or household head A B C D E F G H I J K L M N O P Household type Older single persons or couples (at least one 65+) 7,8 0,379 0,359-5,31 14,94 15,24 0,092 0,085-8,47 13,50 13,62 0,033 0,030-9,86 11,11 11,49 Younger single persons or couples (none 65+) 18 0,238 0,224-5,97 21,68 21,95 0,075 0,070-7,23 25,34 25,92 0,036 0,033-9,28 28,12 29,27 Couple with children up to 18 (no other HH members) 33,6 0,186 0,173-6,67 31,54 31,69 0,050 0,045-9,99 31,39 31,16 0,022 0,019-14,30 31,78 31,26 Mono-parental household 1,5 0,288 0,280-2,83 2,18 2,28 0,088 0,085-4,18 2,48 2,62 0,049 0,043-13,01 3,21 3,20 Other household types 39,1 0,150 0,136-9,71 29,66 28,84 0,037 0,033-11,32 27,30 26,69 0,015 0,013-16,23 25,81 24,81 Socioeconomic group of HH head Blue collar worker 23,3 0,159 0,151-4,81 18,73 19,20 0,034 0,034 0,70 14,79 16,43 0,012 0,012 0,26 12,09 13,91 White collar worker 14,9 0,035 0,035 1,41 2,63 2,87 0,005 0,005-1,35 1,28 1,39 0,001 0,001 4,62 0,61 0,74 Self-employed 23,3 0,243 0,215-11,60 28,64 27,26 0,079 0,065-18,36 34,54 31,09 0,039 0,030-23,28 39,43 34,72 Unemployed 2,3 0,305 0,330 8,12 3,54 4,13 0,083 0,081-2,03 3,57 3,85 0,037 0,037-1,91 3,74 4,20 Pensioner 27,9 0,253 0,233-7,77 35,63 35,38 0,066 0,061-8,18 34,46 34,89 0,026 0,024-10,45 32,03 32,92 Other 8,4 0,259 0,247-4,42 10,98 11,30 0,073 0,072-1,67 11,55 12,52 0,034 0,033-2,91 12,29 13,70 Educational level of HH head Tertiary education 20,4 0,043 0,042-1,28 4,41 4,69 0,009 0, ,13 3,60 20,74 0,003 0,003 1,94 2,80 3,28 Upper secondary education 27 0,152 0,146-3,84 20,74 21,48 0,043 0,040-6,68 21,72 22,35 0,019 0,017-10,50 22,33 22,94 Lower secondary education 13 0,221 0,197-10,90 14,54 13,96 0,053 0,049-7,46 12,95 13,22 0,025 0,022-12,19 13,88 13,98 Primary education or less 39,5 0,302 0,278-7,79 60,18 59,76 0,083 0,074-11,13 61,60 60,36 0,035 0,030-14,55 60,86 59,68 Age of HH member Below ,205 0,192-6,02 27,94 28,28 0,057 0,052-9,30 28,74 28,74 0,025 0,022-13,57 29,53 29, ,5 0,152 0,139-8,62 40,38 39,73 0,041 0,037-10,19 40,47 40,08 0,018 0,016-13,80 42,00 41,55 Over 64 20,6 0,305 0,286-6,24 31,79 32,09 0,080 0,074-8,15 30,89 31,28 0,032 0,028-10,73 28,57 29,27 A: Population Share B and C: Poverty Index FGT0 (distributions of equivalent disposable income before and after the inclusion of imputed income) D: % change in poverty E and F: contributions to aggregate poverty FGT0 (distributions of equivalent disposable income before and after the inclusion of imputed income) G an H: % Poverty Index FGT1 (distributions of equivalent disposable income before and after the inclusion of imputed income) I: % change in poverty, J and K: % contributions to aggregate poverty FGT1(distributions of equivalent disposable income before and after the inclusion of imputed income) L and M: % Poverty Index FGT2 (distributions of equivalent disposable income before and after the inclusion of imputed income) N: % change in poverty, O and P: % contributions to aggregate poverty FGT2(distributions of equivalent disposable income before and after the inclusion of imputed income)