1
Inequality, Poverty and Deprivation in Uruguay (1989-97)
Máximo Rossi (*)
Working Paper. Department of Economics, University of Uruguay
Abstract
The purpose of this paper was to study the evolution of inequality, poverty and deprivation in Uruguay between 1989 and 1997. We found that from 1991 there was an increase wage inequality in Uruguay, poverty changed little, decreased until 1993 and then increased and deprivation increased.
Near a half of poor people in Uruguay are children and old people contribute very little to poverty.
JEL: D300, D630, I320
(*) I thank the research assistant from Tatiana Rossi. Departamento de Economía, Facultad de Ciencias Sociales, Universidad de la República (Uruguay) (). The earlier version of this paper was a Working Paper of the Department of Economics, University of Uruguay, Nº21/2001.
1.-Introduction
The purpose of this paper will be to study the evolution of inequality and poverty in Uruguay between 1989 and 1997.
Uruguay is mainly a urban country. Half of the total urban population lives and nearly two thirds of the economic activity is carried out in the metropolitan area of Montevideo, the capital. The other half of urban population and one third of economic activity are dispersed in the rest of the urban Uruguay (RUC), which includes cities generally not larger than 30,000 inhabitants. Uruguay shows low levels of inequality compared to other Latin American Countries income distribution , and this has not varied too much during the last years. This is in contrast to the situation experienced by the remaining Latin American countries that have increased their levels of inequality.
However, recent studies revealed greater inequalities in some of the components of the households income. Bucheli and Rossi (1994) show important changes in the distribution of pensions; Miles and Rossi (1999); Gradin and Rossi (2000) show a growing inequality in the distribution of wages from the beginning of the 1990s.
The macroeconomic framework in the country can be summarized as follows. After a big recession at the beginning of the eighties, but the Uruguayan economy substantially grew after the recovery of democracy in 1985 until 1994. By 1995 the country went through a new recession that finished in 1996. The period is also characterized by a stabilization plan that reduced inflation considerably, and an increasing opening of Uruguayan economy within the free trade area of MERCOSUR with Argentina and Brazil. A deep reform in the state was conducted but unlike from other Latin-American countries, considerable areas of public intervention were preserved.
The evolution of the distribution of income and poverty in Uruguay is closely related to important transformations in the labor market and in the social protection system.
Regarding the labor market, the country experienced an increase in women’s participation rate as well as in the level of education of the new generations entering the market. A demand bias favoring most skilled people was also observed. Furthermore, this labor market experienced a crucial institutional reform affecting the degree of centralization in wage negotiation. Until 1990 wage increases were decided in bargaining councils by unions, employers and government representatives, and wages adjusted three times a year for all economic sectors and uniformly for Montevidean and RUC workers. A decentralization process begun in 1990, with wage increases decided on a local level and bargaining councils practically disappearing.
Another important change, from the point of view of its consequences in the distribution of income and poverty, took place in the social protection system and is related to the indexation of pensions. Before 1989, pensions were adjusted yearly and linked to the wage index. The reform approved by referendum in December of 1989, established that increases had to take place in the same month as public sector wages (more than one per year) and the rise had to be equivalent to the variation of the wage index within the adjustment period. This fact, in a context of high inflation rates implied substantial improvements in the real level of pensions, moving this group up in the averall distribution of income.
2. The data and privación-inequality-poverty measurement
The study will be based on data from the Household Survey of Uruguay from 1989 through 1997 (Encuesta de Hogares, Instituto Nacional de Estadística). This survey is carried out, in its present format, every month since 1981; its sample framework is the whole civilian population of Uruguay, decomposed in a survey for Montevideo (the capital) and another for the rest of the urban country. It contains individual data on monthly labor earnings, non-labor earnings, age, sex, educational level, hours worked per week, marital status, occupation characteristics, and other relevant variables. All monetary variables will be deflated using the consumer price index of December of 1996.
2.1 Inequality [1]
To measure inequality, I will use three indices consistent with the Lorenz criterion: the Gini coefficient, the Theil index, and the coefficient of variation. If we transfer money from one individual to another with a lower wage, the three indices will register a reduction of the inequality. The main difference between the measures is that if we consider a transfer that reduces the inequality and at the same time and another that increases it, the final result will depend on the weight that each one assigns to both. This weight will depend on the position in the distribution of the affected individuals. The indices show different senstitibity to transfers that take place in different points of the distribution.
Let us consider a group of wages xi, i=1,...,n that have the distribution function F. The mean is . The Gini coefficient G is defined as the area between the actual Lorenz curve and the line of perfect equality. It can be written as:
.
This index is more sensitive to transfers that take place in the center of the distribution, while the coefficient of variation and Theil index are more sensitive to the tails of the distribution. If I denote ln for the logarithm, the Theil index is:
and the coefficient of variation is:
.
It should be kept in mind that the Gini coefficient is bounded between 0 and 1, while the other two measures do not have an upper bound.
2.2 Poverty[2]
The poverty line I will use is a relative one, which will be set at 50% of the median income.
For each individual in the household I compute the equivalent income, defined as the total income of the household divided by the number of individuals in the household corrected by potential economies of scale in consumption. If I denote Yi the income of individual i, the equivalent income (Y´i) is:
Y´i= (Yi ) / (di)
Where the demographic variable di , measures the number of family members and the elasticity, , varies between 0 and 1. I will use four types of equivalent scales: =0.75, =0.55, =0.36 and =0.25. The first assigns the largest increase in cost for increases in family size and gives little weight to potential economies of scale in consumption, whereas the last assigns the greatest economies of scale.
For the dimension of poverty, I will use the index proposed by Foster et al (1984):
where N is the size of the sample, q the number of poor individuals, Z the poverty line and gi = Z - Yi is the poverty gap for individual i, his income being Yi.
The measure P0 is the headcount ratio index: it estimates the percentage of individuals whose equivalent income is below the poverty line. The index calculated with a = 1 weights the headcount ratio by the average of the gap of the poor. Thus the ratio P1/P0 is the average poverty gap among the poor. When a = 2, the index is sensitive to the income distribution among the poor: the wider the poverty gap for individual i, the bigger its weight in the calculation of the index.
One of the advantages of this index is that it is additively decomposable. For each group j of size nj, an index can be calculated:
where gij is the poverty gap for individual i belonging to the group j and qj the number of poor in the group. Thus, Pa is equal to the sum of these measures for every class weighted by the population share nj/N.
2.3 Deprivation [3]
Let us suppose n individuals and that each has a deprivation degree that can be represented by the value of the variable Di ; for higher values of Di bigger person's deprivation.
Di is a function of k linear factors: Xik, k = 1,....,k .
The privation index can be expressed as:
Generally Di is not observable or difficultly observable.
What one can observe is certain categorization of the deprivation level that an individual suffers, for example, I can classify as “not deprived”, “mildly deprived” or “strongly deprived”. A person is considered not deprived if Di=0, mildly deprived if 0 < Di <=Dbar and severely deprived if Di > Dbar, where Dbar is average deprivation index.
I can associate Yi (an ordinal variable) to these different levels.
Then:
Yi=1, si Di<=1
Yi=2, si 1 <=Di<=2
Yi=3, si Di>=2
s are unknown parameters to be estimated together with s.
The probability that Yi equal to 1, 2 y 3 are:
Pr(Zi + i <= 0) or Pr(i <= Zi )
Pr(Zi + i <= ) or Pr(-Zii <= - Zi )
Pr(Zi + i ) or Pr(i - Zi )
where >0.
I estimated the model using a Ordered Logit Model.
The explanatory variables used were: Sex, age, age square, education, region, quantity of members of the household over room higher than 6 (dummy variable), dismissed, pensioner1, pensioner2, discouraged1 and discouraged2.
For the deprivation index I defined different conditions that induce to deprivation.
If there are k conditions that induce to deprivation, k =1,....,K, and Iik is a dummy variable (Iik = 1 if the condition is present and Iik = 0 is the condition is absent) then the individual level of deprivation is:
where is the weigh to the k deprivation condition.
I consider:
where pk represents the frequency with which conditions k occurs, then represent a notion of relative deprivation
It is possible to normalize:
,
where
Then Individual deprivation level is:
Now (Di=0 when none of the conditions are present and Di=1 when all the conditions are present).
The two main problems to build the index are which conditions enter in the index and how they are weighted to build the general index. In this paper for Uruguay the conditions are:
1 House quality.
2 State of conservation of the house.
3 Number of individuals over room bigger than three.
4 Origin of the water.
5 Water installation.
6 Sanitary service.
7 Energy.
8 Heat water.
9 TV color.
10 Refrigerator.
11 Occupant of the house, with or without permission.
12 One unemployed in the household.
13 Rights of medical attention.
The weight associated to a condition that induces to deprivation
is the frequency of that condition. Less frequency correspond to high weight in the index.
Associate to the deprivation index each individual is not deprived, mildly deprived or strongly deprived:
Yi =1 not deprived
Yi =2 mildly deprived
Yi =3 strongly deprived.
3.- Wage inequality
The evolution of the wage distribution is shown in the Figure 1 and Figure 2:
It is observed in Figure 1 and Figure 2 a clear tendency to increased wage inequality in Uruguay. This applies both for Montevideo and the rest of the urban country (RUC), especially since 1991. This growth of inequality is captured by the different indexes, being more important if the sensitibity to transfers is larger in the low line of the distribution. The index of Theil grows 21.6% between 1991 and 1996 in the capital, compared to 11.1% in the case of Gini and 9.6% for CV, and something similar happens in the RUC during 1991-97, 24.9% compared to 10.4% and 17.4% respectively. Starting from inequality levels growth is higher in the capital, except in the case of the variation coefficient, more sensitive to transfers that take place in the high line of the distribution, for this index the inequality grew more in the RUC.
Clearly, there is no single unique factor that could explain the observed increase in wage inequality discussed in the previous section. In what follows I analyze some potential causes that have been used to explain wage inequality in developed countries, such as minimum wage effects, changes in wage-setting institutions and de-unionization of the labor force, increased openness of the economy, shifts of labor demand toward and differences in returns to skills.
3.1- Effect of Changes in the Real Value of the Minimum Wage
The minimum wage sets an explicit floor on the wage distribution, i.e., acting as a backstop for the bottom end of the wage distribution, it should tend to reduce wage dispersion.
In Uruguay the legal minimum real hourly wage fell dramatically during the studied period.
However, the real hourly wage distribution was not pushed downward following the fall of the minimum real hourly wage, i.e., the lower tail of the real hourly wage distribution did not collapse towards the new the minimum wage, as happened, for example, in the USA labor market
Notice that, while for the USA labor market the downward shifts of the lower tail of the wage distribution, due to the fall of the minimum wage, explains some of the increase in wage inequality, this explanation does not seem to apply for Uruguay.
Instead, the increase in wage dispersion seems to respond to shifts of the upper part of the real hourly wage distribution over the full period. That is, the increase in wage dispersion could be attributed to the important wage increments of the most skilled workers. This evidence was more severe in Montevideo than in the RUC, suggesting higher returns to skills in Montevideo. See the evolution of the hourly wage of percentile 90: the relation 90/10 growths mainly for the increase of hourly wage of percentile 90.
3.2- Collective bargaining and de-unionization
During the military regime, unions were declared illegal and wages were adjusted based on price stabilization criteria without any type of bargaining between workers and employers. As a consequence, real wages decreased by nearly 50 percent during the dictatorial period.
Unions were legalized and collective bargaining was reestablished by the first democratic government, in 1985. Until 1990, wage increases were decided in bargaining councils by unions, employers and government representatives, adjusted three times a year and for the entire economic sectors and uniformly for Montevidean and RUC workers. A decentralization process begun in 1990, with wage increases decided on a local level and bargaining councils practically disappearing.
These changes in wage setting bargaining institutions have affected the timing of real hourly wage movements.
Finally, there has been an important de-unionization process in the Uruguayan work force, where membership is not compulsory. While in 1986, four of every ten workers were members of labor unions, in 1997 only one of every ten were. Two facts could explain the observed de-unionization process: the fall in the industry employment, where unions have had more preponderance and the null role of wage bargaining councils beginning in 1990.
3.3- Increasing openness
During the 1990s the Uruguayan economy has experienced a liberalization process that has affected the real wages seriously. With the signing of the MERCOSUR trade treaty in 1990 the increase in trade flows was reoriented towards MERCOSUR countries, i.e., these flows passed 40 to 50 percent of exports and were basically concentrated to Argentina and Brazil. This openness policy had several consequences in the Uruguayan labor market. First, the manufacturing production share on the GDP fell drastically during these years, i.e., from 26 percent to less than a 20 percent. Clearly, Argentinian and Brazilian industries are much more competitive in the manufactured products. One of the policies followed by the Uruguayan manufacturing industry to gain international competitiveness, given the stabilization target of the exchange policies of MERCOSUR countries, was to decrease the real hourly wage increments. Second, there has been a reorientation of the production structure. The fall of the manufacturing share in the GDP was compensated with an increase in the share of service and tourism sector. The share of retail trade, restaurants, hotels and transport, on the GDP, increased from a 16 percent (1986) to a 21 percent (1997) while the number of tourists double during the whole period.
3.4- Changes in relative demand for labor
Clearly, the changes in production structure previously described had affected the labor demand in the urban cities. Although the period considered is relatively small for observing important demand shifts, I can say that: manufacturing employment has fallen in both regions with a substantial increment in retail trade and banking in Montevideo and agricultural and mining in the RUC.
Distinguishing by years of education, the shifts in employment out of manufacturing was into retail trade for least educated and towards baking for most educated.
With respect to changes in occupations, I can observe an increase in professionals, technical and administratives in Montevideo while basically of sales and clerical in the RUC.
Less educated workers have increased their occupation in administratives and sales and clerical while most educated in professionals and technicals and managers.
In general terms, I am suggesting that there has been a shift in demand towards more educated workers for those activities and occupations with a higher expected earning, i.e. professionals, managers or banking.
3.5-Human Capital Returns in Montevideo
I observed that high skill workers had enjoyed much higher wage increments that low skill workers. Also, more educated workers are in industries or occupations with higher expected earnings.
Returns to education are higher in the metropolitan area of Montevideo than in the rest of the urban country. Notice that while in the RUC nearly 40 percent of the workers have six or less years of education, in Montevideo this percentage is only of 25 percent. Also, in Montevideo nearly 20 percent of the workers have more than thirteen years of education, while in the RUC less than 10 percent.
4.- Poverty
a.- Changes of the poverty profile during the period
The evolution of poverty, based in p0, is shown in the Figure 3 and Figure 4: