Income, Inequality and Poverty

Transcription

1 INSTITUTO NACIONAL DE ESTATÍSTICA Income, Inequality and Poverty Regina Soares and Teresa Bago d Uva * Serviço de Estudos e Desenvolvimento Metodológico Departamento de Síntese Económica de Conjuntura Instituto Nacional de Estatística Avenida António José de Almeida Lisboa addresses: regina.soares@ine.pt, teresa.uva@ine.pt June 2002 * The authors wish to thank Paulo Parente for the valuable support. This paper is the result of the participation in the project International Comparisons of Poverty, which consisted of collaborative work between France, several members of the European Union and several Central European Countries. INSEE is currently preparing the edition of a book which collects the studies developed by the participant countries.

2 Contents 1. INTRODUCTION...3 DEMOGRAPHIC AND SOCIO-ECONOMIC BACKGROUND...3 MACROECONOMIC PERFORMANCE CONCEPTS, DATA AND METHODOLOGY DISTRIBUTION OF MONETARY INCOME...22 LABOUR SITUATION OF THE HOUSEHOLDER...24 OTHER CHARACTERISTICS OF THE HOUSEHOLDER...25 AREA OF RESIDENCE, COMPOSITION AND MAIN SOURCE OF INCOME OF THE HOUSEHOLDS INEQUALITY...29 DECOMPOSITION OF INEQUALITY BY GROUPS OF HOUSEHOLDS...30 DECOMPOSITION OF INEQUALITY OF INCOME BY SOURCE OF INCOME POVERTY...34 MONETARY POVERTY...34 Factors associated with monetary poverty of Portuguese households...35 LIVING CONDITIONS POVERTY...39 Analysis of poverty in terms of living conditions...40 SUBJECTIVE POVERTY...41 Analysis of some indicators on subjective poverty...42 THE THREE DIMENSIONS OF POVERTY ECONOMETRIC MODELS FOR POVERTY...47 THE VARIABLES OF THE MODELS...50 THE ESTIMATED LOGIT MODEL FOR MONETARY POVERTY...51 THE ESTIMATED LOGIT MODEL FOR LIVING CONDITIONS POVERTY...55 THE ESTIMATED ORDERED LOGIT MODEL FOR SUBJECTIVE POVERTY...59 COMPARING THE THREE DIMENSIONS OF POVERTY...62 REFERENCES

3 1. Introduction The aim of this paper is to contribute to the debate about welfare, poverty and inequality in Portugal. In societies based on market economies, most essential goods and services are purchased in the market. Therefore, income is an indicator of the living conditions of individuals and households, despite its drawbacks. The use of monetary indicators to characterise living conditions is justified by the fact that they measure the lack of monetary resources to access the market and also because they are easily computable. However, non-monetary indicators are important to complement the analysis of living conditions and social exclusion. Poverty has many dimensions that have to be taken into account. Besides low-income level and material deprivation, poverty is also related to illiteracy, short life, lack of public resources, poor health and various forms of exclusion. This study considers three distinct but related concepts: inequality, poverty and social welfare. The analysis is done for the year This study is organised as follows. The remainder of this section analyses the demographic, socio-economic and macroeconomic situation in Portugal in the last decade. Section 2 describes the most important features of the data and methodologies used. Section 3 is a first and simple approach to the characterisation of the income distribution and identification of groups at risk of low income. Section 4 develops the analysis of income inequality and the factors that contribute to this inequality. Section 5 characterises the poverty situation following three approaches: monetary, living conditions and subjective. Finally, Section 6 presents econometric models for the three types of poverty in order to identify the factors that are associated to a higher risk of poverty. Demographic and socio-economic background The main feature of the recent demographic evolution in Portugal is that the population is getting older. The proportion of young people is decreasing, while the proportion of elder population is increasing. In 1990, 25.2% of the population was under 18 and 13.6% was aged 65 or more; in 1998, these proportions were 21% and 15.2%, respectively. 3

4 From 1975 to 1995, the number of births decreased systematically. In the last five years, it seems that an inversion of this tendency occurred, but only the data for the next years can confirm this as a steady tendency. Nevertheless, since the beginning of the eighties the 2.1 children per couple threshold that ensures the generation replacement is no longer achieved (1.4 in 1998). Life expectancy at birth has been increasing and it was 76.0 years in From 1990 to 2000, the increase in life expectancy for men and women was exactly the same (1.9 years) but the level was 7 years higher for women. Life expectancy Infant mortality Natural increase per Men Women per 1000 of live births 1000 people Source: INE SESP The number of deaths in the last two decades was stable, around one hundred thousand individuals per year. The infant mortality has decreased, in 2000 it was about 52% of the observed in The Portuguese natural demographic balance remained positive during this period. Traditionally, Portugal has been an emigrant country but in the mid-nineties the number of immigrants surpassed the number of emigrants. This happened not only due to an increase of the flow of immigrants from the Portuguese ex-colonies in Africa and from some Eastern European countries, but also due to a reduction of emigration (traditionally to the European countries). Internally, there have been important migrations from the rural communities of the interior to the urban communities of the littoral, mainly of young people seeking better jobs and living conditions. Consequently, the rural population is older and less qualified, as is shown in the next two graphs. 4

5 Population by age Urban Semi-Urban Rural 5 0 <9 10 to to to to to to 69 >69 Source: ECHP 1997 Highest level of education Urban Semi-Urban Rural Total rd level 2nd stage 1st stage Less than 1st stage Source: ECHP 1997 About 86% of the population lives in the regions Norte, Centro and Lisboa e Vale do Tejo. These regions have, respectively, 33.2%, 35.4% and 17.2% of the population. The population in the other regions is 5.2% in Alentejo, 4.2% in Algarve, 2.4% in Açores e 2,5% in Madeira 1. In 2000, the activity rate was 51.1%, with a significant difference between men (57.7%) and women (44.9%). The structure of employment by main activity in different areas shows the importance of the service activities, especially in urban areas. The Services sector is the most important activity in all regions except for Norte, where Industry is the most important sector. As for Agriculture, it is the least important activity in all regions except for Alentejo and Algarve where it is still the second most important activity. 1 Source: Preliminary data from Census

6 Agriculture Industry Services 20 0 Urban Semi-Urban Rural Source: ECHP Agriculture Industry Services 20 0 Norte Centro LVT Alentejo Algarve Açores Madeira Source: ECHP 1997 From 1992 to 2000, the structure of employment by level of education did not change significantly. In 2000, 65% of the employed had not completed the second stage of secondary level education, 14% had completed only this level and 21% had completed the third level. Employment by level of education (year 200) 65% 14% Less than 2nd stage 2nd stage 3rd level Source: LFS 6

7 The definition of unemployment rate changed in Before 1998, people aged 14 years or over were considered in the active population. This age limit changed to 15 years old in Thus, it is quite natural that the results since 1998 differ significantly from the previous years. The next table shows the evolution of the unemployment rates by level of education from 1992 to It can be seen that the unemployment rate increased for the period 1992 to 1996 for all levels of education and that these rates dropped in It is difficult to evaluate the evolutions from 1997 to 1998 due to the change in the definition of the unemployment rate mentioned above. However, it can be concluded that between 1998 and 2000 the unemployment rate has decreased systematically. This is due to a sustained growth of the economy (that is analysed in next section). Unemployment rate by level of education Total Less than first First stage Second stage Recognised third % 3.2% 5.8% 6.0% 3.2% % 4.2% 7.4% 7.0% 4.9% % 5.4% 8.9% 8.6% 6.3% % 5.4% 8.9% 9.8% 6.8% % 5.2% 9.0% 9.6% 7.0% % 4.5% 9.0% 9.0% 6.5% % 3.6% 5.8% 6.2% 5.2% % 3.1% 4.8% 6.1% 4.7% % 2.7% 4.5% 5.6% 3.9% Source: LFS The unemployment affects women more than men and the unemployment rate evolution does not show any indication that they will be closer in the near future. 10 Unemployment rate by sex Men Women Source: LFS Remark: The dashed line marks the date in which there was a change in the definition of unemployment rate. The analysis by age group shows that young people are more affected by unemployment: the group of people under 24 years old has the highest unemployment 7

8 rate, both for men and women. Within groups of age, women are always more affected by unemployment Unemployment rate by sex/age Total- Men M M M >44 M Total- Women W W W >44 W Source: LFS 2000 Macroeconomic Performance Poverty should not be seen as an isolated phenomenon since it is not independent of the economic environment. Indeed, one would expect a worse situation concerning poverty during a recession, when usually there are fewer employment opportunities, smaller growth rates of wages and higher inflation. It is not sought here an empirical evidence of these facts. Nevertheless, it is important to describe the economic situation of Portugal in 1997 in that it sets a relevant background for the analysis of poverty that is done in this study. The economic situation of Portugal in 2000 is also described in order to shed some light on what might be expected on poverty during this year, corresponding to the most recent information available on Portugal 2. The comparison between the economic situation in Portugal in 1997 and in 2000, cannot ignore that these years correspond to different stages of the process of integration of the country in the European Monetary Union (EMU). The selection of countries constituting the EMU was based upon the economic performance of each country in the year 1997, evaluated according to some convergence criteria. Briefly, these criteria were the following: stability of prices, convergence of interest rates, 2 Although there are already some forecasts for 2001 and 2002 provided by the Bank of Portugal, it was chosen not to mention them here. Notice that the terrorist attacks in New York in September 2001 might affect consumers confidence on the USA, which is likely to have some influence on the World economy. 8

9 stability of the exchange rate and fiscal discipline. On the other hand, the year 2000 corresponds to the second year of the EMU since it was implemented on January Apart from the year 1993, the gross domestic product (GDP), evaluated at constant prices, increased during the period For international comparison purposes, next table presents Gross Domestic Product (GDP) real growth, exchange rates and GDP per capita adjusted by Purchasing Power Parities (PPP- prices 1995). The PPPs are the ones published by OECD that uses EKS method for their calculus. Different methods would give different answers in the international comparison context and there is not consensus about the method of aggregation to be used to allow regional or world comparisons. Eurostat and OECD share the responsibility for the calculus of the PPPs, under the Joint OECD-Eurostat PPP Programme. Gross Domestic Product per Capita Year Growth Exchange rates PPP GDP in PPP (PTE/US$) (PTE/ US$) (prices 1995) % % % % % % % % % % Sources: Growth rate- Eurostat (the rate considered here for the year 1995 was the one provided by Quarterly National Accounts- INE) Exchange rate - Federal Reserve Bank of St. Louis and Federal Reserve Board of Governors (USA) PPP- OECD According to the Quarterly Economic Bulletin (March, 1998) of the Bank of Portugal, 1997 was a year with low inflation (1.9% 3 ) and a strong economic growth. Not only did the stability in prices result from a favourable evolution of international prices, but it was also a consequence of the stability of exchange rates and low growth rate of nominal wages. It is worthwhile to notice that the international price of petroleum dropped 7.1% in This is measured by the annual mean variation of the Harmonised Consumer Price Index. 9

10 The Government efforts to control the budged deficit and the public debt were also important in that they brought some credibility to the process of Portuguese integration in the Monetary Union. Since this was foreseen by financial markets, the long-run nominal interest rates decreased. Despite having to keep a tight control of the budget, the Government managed to increase its investment in 12.9%. Main Economic Indicators - Growth Rates Year Private Consumption 3.2% 3.4% 5.1% 4.8% 2.6% Public Consumption 3.4% 2.2% 3.8% 4.5% 2.5% Gross Fix Capital Formation 3.3% 12.0% 12.4% 7.7% 4.2% Internal Demand 3.3% 5.1% 6.6% 5.5% 3.0% Exports 7.1% 7.1% 9.2% 3.2% 8.1% Imports 5.0% 10.0% 14.2% 8.7% 6.0% GDP 3.8% 3.9% 4.5% 3.4% 3.4% Source: Quarterly National Accounts (INE-Portugal) The expansion of the economy during this year is mainly due to the growth of the internal demand. Indeed, the growth rate of this component of the GDP reached 5.1%, reflecting the growth in the Gross fixed Capital Formation (12.0%) and private consumption (3.4%). The main reason behind this positive evolution of investment and private consumption was the reduction of the interest rates. Moreover, employment increased p.p. and there was a rise of real wages. It is also important to note that unemployment rate decreased to 6.7% (it was 7.3% in 1996). However, in this year there was a negative effect on the net exports 5 of goods and services. In fact, while imports increased 10.0%, exports increased only 7.1%. On the other hand, the net transfers increased from 6.3% to 6.6% of the GDP due to an increase of the transfers from the European Union and from the emigrants. Consequently, the Current Account Balance presented a deficit of 1.8% of the GDP. Concerning the year 2000, the GDP had about the same growth than in the previous year (3.4%). In the year 2000 there was a smaller growth of private consumption in 4 This was due to a growth of the number of short-term contracts. 5 This is equal to the difference between exports and imports. 10

11 durable goods and of investment. The overall deficit of the current balance and capital balance was 8.6% of the GDP, which is 2.3 p.p. more than in This was a consequence of the reduction of transfers from the European Union the decrease in net exports (see Quarterly Economic Bulletin of the Bank of Portugal, March, 1998). Notice that the latter was due to a rise of international prices and not to a variation of the exported or imported quantities. In fact, not only was observed a growth of 8.1% of exports, but also a growth of 6.0% of imports. The rise of international prices was related to a rise of the international price of petroleum. The unemployment rate dropped during this year to an annual average of 4.0% (it was 4.4% in 1999) and employment increased 1.7 p.p 6. Despite the inflation rate being 2.9% (which corresponds to a rise of 0.6 p.p. relative to 1999), the growth of real wages was only of 2.1% (2.4% in 1999). The goods that contributed the most to this inflation level were the foodstuff group and the fuel group. Notice that during this year the European Central Bank increased the interest rate six times in order to reduce the pressure on prices due to excessive demand. The deceleration of the internal demand was required since the level of debt of families and companies had been too high before In what concerns the public budget, it is worthwhile to notice that its control was achieved mainly with the reduction of direct investment. In fact, the public consumption increased less than in To sum up, not only the economic growth was higher in 1997 than in 2000 but also the inflation rate was lower in Additionally, the interest rates had different trends. That is, while in 1997 the interest rates decreased, in 2000 the interest rates increased. Although most of the economic indicators unveil a better economic performance of the year 1997, there is one exception. The growth rate of wages of the public employees was 3.7% in 2000 and it was 2.75% in However, this was a consequence of a commitment of the Government to compensate the workers for the high inflation of the previous year. 6 It should be mentioned that the unemployment rate of 1997 cannot be compared with the unemployment rate of Since 1998 only people with a minimum age of 15 years old have been considered to compute the active population. The comparison of the unemployment rates of 1997 and 2000 will be misleading in that before 1998 this minimum age had been 14 years old. 11

12 2. Concepts, data and methodology The data used in this study were taken from the fourth Wave, year 1997, of the European Community Household Panel (ECHP). This is an annual survey carried out by most of the Community member states, with common methodologies to ensure comparability. The ECHP is nationally and regionally representative and aims to provide a reliable portrait of socio-economic conditions across the countries. The questions are asked regarding both households and individuals within the household and cover different topics including demographic variables, characteristics of the dwelling, education, health, employment and income from different sources. A criticism can be done to this data source: some of the poorest people are not considered in the statistics. In fact, the panel includes only people living in households, hence excludes those living in institutions and homeless. This limitation can not be forgotten when analysing results but, still, the ECHP seems to be the best source of information available. The definition of household is established in terms of two criteria: sharing accommodation and having common living arrangements. There is no single definition for the reference person of the household in the ECHP. In Portugal, the reference person was self-selected by the household, which does not correspond to an objective characterisation. Thus, in this study, the reference person of the household - the householder - is the person whose contribution for the budget is the most important (although this is an arbitrary definition, it is reasonable in economic terms). The ECHP has information on income by individual and by household, and, in both cases, information on different components of the total income. The reference year for the income is the calendar year before the survey, thus, 1996, in the 1997 wave used in this study. The income concept used in ECHP is net monetary income, it does not take into account non-monetary income, like savings from consuming goods from own agriculture, own business or home production. Net monetary income underestimates real income for the households that also have non-monetary income. This is a matter of concern because, for a great share of households, this component is a significant 12

13 proportion of total income. According to Households Budget Survey 2000, nonmonetary income represented 15% of total income. The choice of income unit is relevant in the analysis of poverty and inequality. Total income of the household does not take into account the composition of the household. The household income per capita only takes into account the size but not the scale effects on income use. In this study, the household income per equivalent adult is assigned to the individuals, and the individual is the unit of analysis. Thus, account is taken of the different needs of the households according to their size and composition, as well as the economies of scale that occur from sharing fixed housing or other fixed costs. There are various approaches to derive an equivalence scale 7. The choice of the equivalence scale affects the distribution of monetary income (expenditure), thus, the inequality and poverty indices. Inequality tends to be smaller when incomes are adjusted, but this also depends on the sensitivity of inequality measures to the various parts of income distribution. The analysis is sensitive to the definition of needs implicit on the definition of the equivalence scale. The equivalence scale used in this study is the OECD modified scale which gives the weights 1, 0.5, and 0.3, respectively to the first adult, the remaining adults and the children under 16 years old. Parametric scales are often criticised because of the arbitrary choice of parameters but, in spite of their simplicity, they can approximate the results given by other more sophisticated methods (Medina, 1999). Poverty has many dimensions, therefore, the measurement and analysis of poverty should use a set of indicators in order to reflect those different dimensions. The analysis of poverty usually follows the monetary approach, which only takes into account the income or consumption levels. According to this approach, a household is considered to be poor if the income (consumption) is below a threshold, called poverty line. The monetary poverty line can be defined in an absolute way as the level of income considered enough to satisfy the consumption needs of a family type or in a relative way as a function of the overall distribution of income. 7 A survey of the various approaches is given in Coulter, Cowel and Jenkins (1992). 13

14 Absolute poverty was defined by the United Nations (1995, page 57) as: A condition characterised by severe deprivation of basic human needs, including food, safe drinking water, sanitation facilities, health, shelter, education and information. It depends not only on income but also on access to services. Atkinson (1998, page 21) defines absolute and relative poverty lines in the following way: An absolute poverty line is usually taken to be one which is fixed over time in terms of purchasing power, allowing the purchase of a specified basket of goods and services which has some justification as the minimum necessary or basic needs. A relative poverty line varies with changes in average income; it can be defined as a fixed proportion of the average. The basic needs differ from country to country and with time, according to the level of development, culture, etc. There is always a degree of arbitrariness in the definition of the basket of goods and services necessary to satisfy the basic needs. Moreover, the ECHP does not contain information on household expenditure, so it is not possible to establish an absolute line. Then, the monetary poverty line used in this study follows the relative approach and is set as 60% of median annual income per equivalent adult. The median income is chosen instead of the mean income because this is very sensitive to changes in the tails of the distribution. The median is unaffected by the extreme values, but it is affected by the deletion of observations, such as the deletion of zero income observations. The choice between mean and median income is not only a statistical issue and affects the level of the poverty line. In Portugal, the median is 80% of the mean (60% of the median corresponds to 48% of the mean), therefore, to define the poverty line as a function of the mean instead of the median would lead to a higher estimate for the number of poor people. There is no consensus about the percentage to choose to define the threshold and a different choice would lead to a different estimate of the poverty rate. In this study, the choice took into account the minimum national salary, the guaranteed minimum income and some social benefits. Given the monetary poverty line (z), different poverty indices can be defined, which aggregate in different ways information of those below the threshold. 14

15 The most common poverty measures belong to a class proposed by Foster, Greer and Thord (1994): α 1 = n where y i is the income of individual i. P 0 is the Headcount Index. P p i= 1 z y max z i,0 P 1 is the Poverty Gap, it takes into account how far the poor are from the poverty line. P 2 is the Poverty Intensity, it incorporates some convexity to the distances between yi and z and is sensitive to inequality among poor. There is not always a clear relationship between lack of monetary resources and lack of essential commodities in order to live according to the standards of the society. Moreover, the limitations of the variable used to estimate monetary resources give additional reasons to consider more than just this variable. Therefore, apart from the monetary approach, some complementary approaches should be used in order to analyse the different dimensions of poverty. The living conditions poverty can be studied based on a set of goods and services considered representative indicators of the standard of living of society. The choice of a methodology to derive a poverty line according to this approach is far from being consensual since it is not easy to agree upon the indicators that should be taken into account. The setting of a poverty line in terms of living conditions can follow an absolute or relative standard (as was explained above for the monetary poverty line). An absolute poverty line can be defined using a set of living and housing conditions considered essential. One that selects the households that bear the worst conditions, whether or not they have those considered essential, is a relative poverty line. This study uses an approach similar to the one used in Lollivier and Verger (1997). Households are classified according to a score built as the sum of various indicators of poor living conditions (Branco, 2002). The poverty line is a value of the score so that the resulting poverty rate equals, as much as possible, the monetary poverty rate. This poverty line is a relative one because a fixed proportion of the households that bear the worst housing and living conditions is selected. α 15

16 A different approach of poverty can incorporate some dimensions that are not considered by the previous indicators. This approach is called subjective poverty and considers the way people perceive their own living conditions, financial situation, health, isolation, social exclusion. This approach recognises the importance of the subjective judgement of people about their standard of living. Income poverty influences subjective feelings on poverty. Subjective poverty lines can be based on the opinion of people regarding the minimum income necessary for the household to make ends meet. A poverty line based on this opinion might lead to some inconsistencies since people with the same economic welfare may give different answers (Ravallion, M. 1998). Pradhan and Ravallion (1997) proposed another method based on qualitative data on consumption adequacy. In this study, the subjective poverty indicator is derived from the opinion of the household about the ability of making ends meet (see Branco, 2002, for details on this indicator as well as the analysis of alternative indicators for subjective poverty). The extent of the subjective poverty is defined as the difficulty to make ends meet. An indicator of the subjective poverty is observed. It takes values from one, if the household is able to make ends meet very easily, to six if that is very difficult. The subjective poverty line was defined considering as poor the households that have difficulty or great difficulty to make ends meet. Both the living conditions and the subjective approaches result in an indicator of poverty. These indicators can be aggregated using the P 0 measure defined above, leading to the living conditions poverty rate and the subjective poverty rate. Different approaches to set a poverty line lead to different rates and different profiles of poverty. The importance of following diverse approaches is that it is possible to put into action different policies to fight poverty according to those poverty group profiles. Concerning the analysis of poverty, this study follows the three approaches explained above: monetary, subjective and living conditions. Several inequality indicators have been proposed in the literature. It is desirable that a set of axioms is satisfied: 1. Pigon-Dalton Transfer Principle the index does not decrease in response to a transfer from someone to a richer person and it does not increase when an income transfer occurs from someone to a poorer person. 2. Income Scale Independence the index is invariant to uniform proportional changes. 16

17 3. Principle of Population the index is invariant to replications of the population. 4. Symmetry the index is independent of any characteristic of the individual other than income. 5. Decomposability inequality of the overall distribution is consistent with inequality of parts of the distribution, such as population sub-groups. Many measures satisfy these conditions. Let y i be the income of individual i, µ ( y) be the arithmetic mean of the distribution of income, and n the number of individuals. The Gini Coefficient (G) is defined as: 1 G = 2 2n µ n n ( ) y i= 1 j= 1 y i y j The Gini coefficient fails the decomposability axiom but allows other useful, though less desirable, decompositions. The Generalised Entropy Measures (GE) generally meet the set of desirable axioms. They are defined as: GE ( α) In this class of measures, the most used are: Mean Log Deviation 1 GE(0) = n Theil Index α n 1 1 y ( ) i = α 2 α n i= 1 µ y 1 n i= 1 n 1 y GE(1) = n µ µ log y () y i i i log ( y) µ ( ) y i= 1 y Transform of the Coefficient of Variation n 1 GE ( 2) = µ 2nµ ( y) 2 i= 1 ( ( ) 2 y i y 17

18 Another class of measures is the Atkinson Measures that are defined as: A ε 1 ε n 1 y = µ ( ) i = 1 n i 1 y where ε is the inequality aversion (high values of ε mean high inequality aversion) and 0 < ε <. Ae takes on values between 0 and 1. If measures for α < 1. 1/ ( 1 ε ) α = 1 ε the GE measures are equivalent to Ae These measures of inequality have different sensitivities to transfers in the different parts of the distribution and this may lead to order the same two distributions in opposite ways. The Gini index is more sensitive to transfers in the middle of the distribution and takes on values between 0 and 1. It may take negative values if mean income is negative or a value greater than one if there are very large negative incomes, but zero incomes are no problem 8. The value of GE measures ranges from 0 to. A value of zero represents an equal distribution of income. Higher values represent higher levels of inequality. In the GE measures, α represents the weight given to distances in the different parts of the distribution. For α = 0, GE gives more weight to distances between incomes in the lower tail of the distribution; α = 1, means that equal weights are given to distances between income across the distribution; with α = 2, the distances between income in the upper tail of the distribution are given more weight. GE(0) will always reach its maximum in the presence of zero incomes. When there are negative incomes, α must be greater than one so that GE(α) is defined. The sensitiveness of the Atkinson index depends on the value of ε: the higher the values of ε, the higher the weight given to lower incomes. A 1 will always attain its maximum in the presence of any zero incomes. Decomposability is a desirable propriety for measures of inequality because it allows the analysis of the relation between a characteristic and the inequality of income. The decomposition of inequality by subgroups (resulting from dividing the population 8 Scott and Litchfield (1994) 18

19 according to a given characteristic) identifies how inequality within each subgroup and inequality between different subgroups contribute to the overall inequality. Some household and personal characteristics, such as gender, education attainment, occupation, age, location of the household, sources of income of the household, are determinants of household income. The inequality decomposition can give the contribution of these factors for total inequality. This contribution is given by the between-group inequality and reveals the inequality between people with different characteristics. For any partition of the population some inequality exists among each subgroup. This heterogeneity inside the subgroups is given by the within-group component of inequality. Decomposition of inequality by population sub-group The GE measures of inequality can be decomposed in a very simple form. Let Π () k be a partition of the population into k subgroups, indexed by j. The between-group component is defined as: I B K 1 µ = f j α α j= 1 µ α ( y) ( ) j y 2 1, where f subgroup j. n j j = is the population share of subgroup j and ( y j ) n The within-group component is defined as: where w j = v α j f 1 a j and I w ( y) ( y) Then, the total inequality is defined as 9 : v j = k j= 1 w GE j () a n j µ j = is the income share of group j. nµ I = I B + I w. j, µ is the mean income in I I B w gives the share of inequality explained by a given partition of population and depends on the GE measure being decomposed and on the attribute of the population that defines the partition. 9 Cowell and Jenkins (1995) 19

20 Inequality decomposition can also be dynamic in order to analyse changes in inequality over time. Dynamic Decomposition The decomposition (according to a partition of the population) of the variation of inequality between two periods must take into account different causes for the changes. These causes are variations in the size of subgroups, changes in relative income between subgroups and changes in inequality within groups 10. Mookerjee and Shorrocks (1982) have shown that the overall change in GE(0) can be well approximated by the following expression: k k () = GE() 0 j f j + λ j log( λ j ) j= 1 j= 1 k [ ] f + ( v f ) log ( y) GE 0 µ j j= 1 j j k ( j ) + f j GE() 0 where the overbar means the average of the values of the variable at t and t+1 ( ) ( y) λ = is the ratio of subgroup mean to overall mean and is the difference j µ y j µ operator. The first and the second terms of the expression give the population shares effects, the third term is the relative mean income effect and the last term is the change of inequality within groups. A negative sign in any component indicates that the respective term contributed to the decline in inequality measured (according to this index). Decomposition of income Total income results from more than one source. Those components of income (labour earnings, income from capital, etc.) are likely to have different contributions to total inequality. Many of the usual measures of inequality are undefined for negative incomes, therefore, do not allow a breakdown by component of income because some components may be negative. For example, incomes may contain losses from self employed or unincorporated business activities or it may be interesting to consider as negative incomes the charges supported by individuals (taxes, social security payments, etc.) to measure their effect in inequality. In the GE class, the only measure that is defined for negative incomes is GE(2). j= 1 j 10 Mookerjee and Shorrocks (1982) 20

21 Total inequality, I, can be expressed as the sum of the contributions of each source of income, f: I = f S f i= 1 where S f is the absolute contribution of the source of income f to overall inequality. The proportional contributions are given by: s f S = I GE(2) is decomposed in the following way: f f f f and s = 1. f f i= 1 S = s GE( 2) = ρ χ GE(2) * GE(2), where ρ f is the correlation between component f and total income and f s factor share, the ratio of mean income of source f to overall mean income., f f µ ( y) f χ f = is µ ( y) A large value of S f suggests that component f of income is an important source of total inequality. If S f < 0, the income source has an equalising effect. The major drawback of these measures of inequality is that, depending on the measure of inequality being decomposed, the importance of the attribute in analysis will be different. There are alternative approaches for the calculation the importance of attributes in explaining the level of inequality that are not considered in this study. Fields (1997) proposes a method that involves running a set of regressions and the estimation of the relative contributions as functions of the estimated parameters. The quantile regression allows the estimation of how the effects of several factors vary for different percentiles of the income distribution and takes into account the heteroskedasticity of the different groups Deaton (1977). 21

22 3. Distribution of monetary income This section introduces the characterisation of the income distribution. The income considered is the annual net household income per equivalent adult, as defined in Section 2. The computation of the income means takes into account not only the household weights, but also the size of the household. This procedure is equivalent to attribute the income per adult equivalent of each household to all its members and to calculate the income means among the individuals. The mean annual income per equivalent adult is $ (9 737 PPP), and the median is $ (7 767 PPP). The large difference between the mean and median income is an indication of the skewness of the distribution of income. The table below shows the mean income by decile. Net mean Income per Equivalent Adult Deciles Escudos PPP 1 st Less than nd to rd to th to th to th to th to th to th to th More than Total The first three deciles of the distribution have a very low level of income (on average, about 22%, 39% and 50%, respectively, of the overall mean income). On the other hand, the mean income for the 10 th decile is almost three times the overall mean income. These values indicate a great dispersion on income that can be confirmed by the analysis of inequality, presented in the next Section 4. The table below shows how the income varies when people are grouped according to their opinion about their financial situation. 22

23 Net mean Income per Equivalent Adult The household is able to make ends meet: Escudos PPP 1- With great difficulty With difficulty With some difficulty Fairly easily Easily Very easily About 40.4% of the population belongs to households that claim to have difficulty in making ends meet. As was to be expected the income level is negatively correlated with the perceived difficulty in making ends meet. The distribution of income, estimated by a non-parametric methodology 12, also shows the asymmetry of the income distribution: Estimated density of the Income per equivalent adult 9,E-07 6,E-07 3,E-07 0,E+00 0,E+00 2,E+06 4,E+06 6,E+06 8,E+06 1,E+07 1,E+07 1,E+07 Income per equivalent adult This asymmetry can be associated with several factors. The following sections analyse how the income varies with some characteristics of the householders and the area of residence and composition of the household. 12 Kernel density estimator - see Silverman (1976) 23

24 Labour situation of the householder The employment status of the householder may contribute to the asymmetry of income. For example, the households whose reference person is inactive or unemployed have, on average, a lower income. Net mean Income per Equivalent Adult Main activity status of the householder Escudos PPP Normally working Unemployed Inactive Among the household whose householder is working, the industry and the occupation are important factors that determine the level of income. If households are classified according to the current industry of the householder, the lowest mean income is observed for the activity Agriculture, hunting and forestry + Fishing and the highest for Financial Intermediation. Net mean Income per Equivalent Adult Current industry of the householder Escudos PPP J Financial intermediation K Real estate, renting and business activities N Health and social work M Education E Electricity, gas and water supply L Public administration & defense; compulsory social security O-Q Other community, social & personal service activities C Mining and quarrying I Transport, storage and communication DF-DI Manufacture of coke, refined petroleum/chemicals/rubber DJ+DK Manufacture of metal products, machinery & equipment n.e.c G Wholesale & retail trade; repair of motor vehicles, H Hotels and restaurants DB+DC Manufacture of textiles, clothing and leather products DA Manufacture of food products, beverages and tobacco DD+DE Manufacture of wood & paper products; publishing & printing F Construction DL-DN Other manufacturing A+B Agriculture, hunting and forestry + Fishing

25 Net mean Income per Equivalent Adult Current industry of the householder Escudos PPP Agriculture Industry Services In general, the lowest mean income is observed in households whose reference person works in Services, followed by Industry. The current occupation of the householder is another determining factor of income. If the householder is a skilled agricultural or fishery worker, the income of his household is, on average, about 52% of the overall mean income, which is about 22% of the mean income of households whose householder is a professional. Net mean Income per Equivalent Adult Current occupation of the householder Escudos PPP Skilled agricultural and fishery workers Elementary occupations Craft and related trades workers Not working Service workers and shop and market sales workers Plant and machine operators and assemblers Clerks Legislators, senior officials and managers Technicians and associate professionals Professionals Other characteristics of the householder The impact of the educational level attained by the householder on the level of income of a household is very clear. If the householder has a Recognised third level degree, the income of the household is, on average, about three times the mean income of the households whose householder did not achieve the Second stage of secondary level and two times the mean income of those that achieved this level. Net mean Income per Equivalent Adult Highest level of completed education of the householder Escudos PPP Recognised third level education (ISCED 5-7) Second stage of secondary level education (ISCED 3) Less than second stage of secondary education (ISCED 0-2)

26 The households whose reference person is a man have, on average, a higher mean income but the difference is not very significant. Net mean Income per Equivalent Adult Sex of the householder Escudos PPP Male Female Classifying the households according to the age of the householder, the lowest mean incomes are observed for the younger and older ages. Net mean Income per Equivalent Adult Age of the householder Escudos PPP Less than From 25 to From 30 to From 35 to From 40 to From 45 to From 50 to From 55 to From 60 to More than The class 50 to 54 years old has the highest mean income. The mean age of the householder is higher in the first deciles, suggesting that older people have the lowest incomes Average age of the householder by decile of Income st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 26

27 Area of residence, composition and main source of income of the households The area of residence of the household also influences the income. Madeira is the region of Portugal with lowest mean income (about 66% of the overall mean income and 52% of mean income of Lisboa e Vale do Tejo). Lisboa e Vale do Tejo has the highest mean income, followed by region Norte, but with a significant difference between them. The mean income of region Norte by equivalent adult is about 70% of the mean income of Lisboa e Vale do Tejo. Net mean Income per Equivalent Adult Regions Escudos PPP Norte Centro Lisboa e Vale do Tejo Alentejo Algarve Açores Madeira There are also significant differences between areas with different degrees of urbanisation. The mean income of households in urban areas is almost double the mean income of rural areas. Net mean Income per Equivalent Adult Degree of urbanisation Escudos PPP Urban Semi-Urban Rural The composition of the household also affects the distribution of equivalent income. Women aged 65 or more living alone have a very low income. The same is observed for households composed by 2 adults with 3 or more dependent children and for single parents. 27

28 Net mean Income per Equivalent Adult Type of Household Escudos PPP 1-person household: Female aged 65 or more adults with 3 or more dependent children Single parents with 1+ dependent child Other household with dependent children person household: Female under adults without depend. child with at least one person >= person household: Male aged 65 or more Other household without dependent children adults with 1 dependent child adults with 2 dependent children adults without depend. child with both under person household: Male under The group of households whose main source of income is investments, savings or insurance have the highest mean income, followed by the group of households whose main source of income is wages and salaries. The households whose income is composed mainly of social transfers have the lowest income, on average. Net mean Income per Equivalent Adult Main source of income: Escudos PPP Wages and salaries Income from self-employment or farming Pensions Unemployment benefits Any other social benefits or grants Income from investment, savings, insurance

29 4. Inequality Inequality can be analysed in several different ways. The analysis of the shares of income in population sub-groups is a simple way of characterising inequality and defining risk groups. Additionally, the analysis can incorporate some social aspects considered by a particular society as important and possible to achieve. In this study, inequality is represented as the dispersion of the distribution of income. The characterisation of this dispersion can be done using various measures as was explained in Section 2. The Lorenz curve shows the partition of the income. In a society with a perfectly equal income distribution, the cumulative share of income equal the cumulative population share across the whole distribution. This is represented in the graph by the 45 line. The gap between the actual line and this mythical line indicates the degree of inequality. The graph below shows that the inequality is high in Portugal. 100 Lorenz curve Next table presents the net mean income shares by equivalent adult and the accumulated income shares in each decile (co-ordinates of Lorenz curve). Shares of Net Income per Equivalent Adult by deciles Deciles 1 st 2 nd 3 rd 4 th 5 th 6 th 7 th 8 th 9 th 10 th Income Shares 0,022 0,037 0,053 0,061 0,074 0,086 0,099 0,119 0,159 0,290 Cumulative Income Shares (co-ordinates of Lorenz curve) 0,022 0,059 0,112 0,173 0,247 0,334 0,432 0,551 0,710 1,000 29