TECHNICAL ANNEX II

An alternative method for identifying characteristics
of poor and non-poor families

As it was shown in Chapter 3, the lowering of the poverty line is of dubious effectiveness as far as identifying the characteristics of poor families concern. In Georgia, low poverty lines hardly provide us with a better understanding of what characteristics distinguish poor and non-poor Georgians. This annex explores the use of an alternative method to identify those characteristics.

The objective is to explore whether families can be grouped in terms of similar levels of welfare. We attempt to identify different groups of families in the Georgian society using some measure of standard of living and then observe what characteristics these groups posses. A potential technique for this purpose is cluster analysis. In brief, the technique identifies different groups (e.g. families) based on their similarity with respect to a given variable. In our case, the variable is consumption over the poverty line.

Cluster % of families Cluster % of families 1 0.46 6 7.74 2 16.77 7 0.09 3 35.12 8 0.18 4 0.28 9 2.12 5 36.59 10 0.65

The initial target was the construction of a maximum of 10 clusters or groups. Table A2.1 above shows that there are four clusters which together account for 96% of all families (2, 3, 5,and 6) Two out of these four clusters, numbered 3 and 5, account for more than 70% of all families. The next table presents the center of clusters and their range. The center is the average value of consumption over poverty line and the range indicates the minimum and maximum within each cluster. Table A2.2 summarizes this information for cluster 2, 3, 5 and 6.

Centers of clusters 5 (poor) 3 (poor, middle class) 2 (relatively well off) 6 (relatively rich) 0.62 1.29 2.19 3.53 Dispersion within cluster Min Max Min Max Min Max Min Max 0.00 0.94 0.95 1.73 1.74 2.83 2.90 4.27

The group of poor people constitutes almost 37% of the population and its consumption is on average 62% of the NHDR Poverty Line. It has a lower bound close to zero and an bound is 95% of the poverty line and it means that some families in the group do not manage to get over the NHDR poverty threshold. The upper limit is 1.73 times the value of the poverty line. The next cluster, which has been called the relatively well off, has a center cluster of more than twice the poverty line and includes 17% of the population. Both its minimum and maximum values are above the poverty line meaning that all families are non-poor. The final cluster accounts for less than 8% of the population and comprises relatively rich Georgians with an average consumption of 3.5 times the NHDR Poverty Line. Both minimum and maximum are comfortably above the poverty threshold.

We now present several tables with characteristics for these four clusters. A total of more than 50 variables were originally explored. However, for reasons of space, the tables below are restricted to spending, employment, a short set of education variables, families perceptions of their own social situation, and expectations about the future.

5 3 2 6 Spending per capita 37.35 75.95 129.20 213.09

Category 5 3 2 6 Row % Col % Row % Col % Row % Col % Row % Col % Unemployed 57.8 39.3 30.7 21.8 8.9 14.4 2.6 8.3 Uncompleted Secondary 64.3 2.3 21.4 0.8 7.1 0.6 7.1 1.2 Technical 34.0 16.1 42.0 20.7 18.6 21.0 5.3 11.9 Incomplete higher 26.7 2.0 60.0 4.7 3.3 0.6 10.0 3.6 Higher institute 26.1 22.9 38.7 35.4 21.8 45.5 13.5 56.0 Post graduate courses 14.3 0.3 28.6 0.5 28.6 1.2 28.6 2.4

Table A2.4 produces no surprises and is in line with the findings of Chapter Three. The cluster of poor people contributes with almost 60% of the unemployed while the cluster of rich people contributes with only 2.6%. The education of the head of household provides arguments for the idea that education should matter. One can notice that the presence of members with higher education characterize rich families. However, the modest contribution of higher education to explain variations in family welfare, as shown in Technical Annex 1, suggest that the well-off in Georgian society are able to send their children to university, and therefore their families are characterized by a greater share of graduates, but that university per-se may not be a determinant in achieving high levels of income. If this line of argumentation has some substance, then the access to those networks that come together with rich families may play a greater role than education as an effective anti-poverty strategy.

Postgraduate courses, which indicate a tertiary degree and further education, are less predominant at lower levels of society though a relatively low percentage of rich people hold one (2.4%). When one takes the group of people with a degree from a higher institute, the majority of members come from poor and middle class. More than half of rich people also hold the same degree but the same proportion is one quarter for the group of poor citizens.

Table A2.5 shows employment categories for each cluster and in general it confirms the findings of the set of regressions on consumption and income. The private sector, with all its problems, confers better capacity to stay afloat. Two thirds of relatively rich people have a family member working in the private sector against only one third in the poorest group. Things start to level down when the employment category is government office. There is an almost equally proportional representation of the middle class and the well off in the rank of government employees. Clusters 3, 2, and 6 have about 40% of their families with at least a member working for a state office. Only the poorest cluster is underrepresented with a quarter of families working in the same category.

The state enterprise offers no great differences with a few percentage points difference among groups. This reflects the fact that state enterprises are a mix bag and much depend on the particular enterprise and more importantly, the position within the enterprise. 25% of relatively well off families has at least one member that is a high rank government officialsss, director or manager. In contrast, only 3% of poor people enjoy the presence of such a member in the family.

Category   5 3 2 6 Row % Col % Row % Col % Row % Col % Row % Col % Family member in private sector No 45.5 65.0 35.8 53.3 13.6 46.1 5.1 34.5 Yes 16.4 35.0 22.7 46.7 40.4 52.7 20.4 63.1 Family members working in government office No 43.8 73.3 33.8 59.1 14.9 59.3 7.5 59.5 Yes 29.8 26.7 28.3 40.9 21.6 40.7 20.3 40.5 Family members working in state enterprise No 38.5 92.7 37.1 93.2 16.4 94.0 7.9 90.5 Yes 29.4 7.3 45.6 6.8 15.5 6.0 9.5 9.5 Family member that is high rank official, Directors, Managers No 40.3 97.0 37.4 94.0 15.8 90.4 6.5 73.8 Yes 6.0 3.0 11.4 6.0 47.0 9.6 35.6 26.2 Presence of service employee No 39.8 85.9 36.1 81.1 16.4 83.8 7.7 78.6 Yes 22.0 14.1 27.9 18.9 27.2 16.2 22.8 21.4 Presence of mid-managers, professional workers No 42.2 88.9 35.6 78.2 14.6 73.1 7.5 75.0 Yes 15.7 11.1 37.5 21.8 28.7 26.9 18.1 25.0 Presence of technical skilled workers No 40.1 89.2 36.1 83.7 15.9 83.8 7.9 83.3 Yes 29.1 10.8 36.7 16.3 21.4 16.2 12.8 16.7 Presence of unskilled workers No 38.4 83.4 36.3 82.2 16.7 86.2 8.7 89.3 Yes 29.8 16.6 50.7 17.8 16.1 13.8 3.4 10.7

This proportion increases linearly up to the relatively well off group (cluster 2) and then almost triples in the cluster of richest Georgians (cluster 6) testifying to the success of this anti poverty strategy. In turn, the presence of a mid manager or a technical skilled worker in the family confers some ability to stay above the poverty line but does not go further than that. Clusters 2 and 6 have a similar proportion of families with a mid-manager or a technical skilled worker. Finally, the presence of unskilled members predominates among the poor and middle class but there is not significant difference between the bottom and the middle of society. Their presence diminishes by 4 percentage points in the group of relatively well off and by another 3 percentage points in the group of rich people.

Category   5 3 2 6 Row % Col % Row % Col % Row % Col % Row % Col % Additional house/apartment No 41.3 86.9 37.6 82.4 14.7 73.7 6.3 63.1 Yes 26.8 13.1 34.5 17.6 22.7 26.3 16.0 36.9 Family has a house in countryside No 39.9 98.2 37.2 95.5 15.8 92.8 7.1 82.1 Yes 13.7 1.8 33.3 4.5 23.5 7.2 29.4 17.9 Member with insurance No 38.9 98.7 37.3 98.7 16.1 97.0 7.8 94.0 Yes 25.0 1.3 25.0 1.3 25.0 3.0 25.0 6.0 Family savings No 40.2 98.0 36.8 93.4 15.8 91.6 7.2 83.3 Yes 13.1 2.0 41.0 6.6 23.0 8.4 23.0 16.7

The presence of an additional house and a presence of a house in the countryside occur in all clusters but not surprisingly it is the relatively rich group the one to that owns more. Almost 40% of relatively rich people own an additional house while 20% posses one in the countryside. It is interesting that the group of relatively rich people is still mostly uninsured with only 6% of them having some type of insurance. It serves as a warning sign about the assumption that 5 GEL a month could cover basic cost of health care if people choose to buy insurance (see Chapter 2). While the assumption is technically correct (the insurance exists, at least in Tbilisi), even rich people choose to go uninsured. The second surprise is the low level of saving of the relatively rich. Only 20% reported having savings. While it is clear that underreporting occurs, 20% is still a low percentage of people with savings.

Table A2.7 summarizes people’s satisfaction with their current job, their assessment of past economic situation and expectations about the future. In general, satisfaction with the present job improves with greater income, but it is noteworthy that more than half of poor people reported to feel satisfied, which is a big percentage considering the low level of income in the group. Clusters 3 and 2, which capture the middle strata of society do not present significant differences. Approximately 2/3 of their members reported that they feel satisfied with their job though roughly half is looking for a change.

5 3 2 6 Row % Col % Row % Col % Row % Col % Row % Col % Satisfied with job position and/or looking for another job Yes, satisfied 20.5 14.9 39.7 26.2 24.0 35.0 15.7 44.4 No, but not looking 38.8 42.2 39.4 38.9 15.7 34.4 6.1 25.9 Yes, but looking 41.2 42.9 36.9 34.9 14.6 30.6 7.3 29.6 How did family economical condition change during the last season? Improved a lot 22.2 0.5 33.3 0.8     44.4 4.8 Improved a little 20.8 5.1 37.5 9.5 30.2 17.5 11.5 13.1 No change 36.6 40.9 42.3 48.9 14.3 38.0 6.8 35.7 Worsened a little 39.5 30.2 34.6 27.4 16.3 29.5 9.6 34.5 Worsened a lot 51.7 23.4 28.7 13.4 14.0 15.1 5.6 11.9 How did family economical condition change during the last year? Improved a lot 11.1 0.3 55.6 1.3     33.3 3.6 Improved a little 23.6 7.7 37.8 12.7 26.8 20.5 11.8 18.1 No change 36.4 34.5 41.0 40.2 16.7 37.3 5.9 26.5 Worsened a little 39.4 32.0 37.9 31.7 13.2 25.3 9.5 36.1 Worsened a lot 51.5 25.6 27.3 14.0 14.4 16.9 6.7 15.7 How will family economic conditions change within the next year? Improve a lot     54.5 2.1 18.2 1.5 27.3 4.6 Improve a little 26.7 18.1 37.1 25.7 21.8 32.8 14.4 44.6 Remain the same 35.7 38.8 42.2 46.9 16.0 38.8 6.2 30.8 Worsen a little 48.4 24.7 33.3 17.5 13.1 14.9 5.2 12.3 Worsen a lot 55.6 18.4 23.2 7.9 16.2 11.9 5.1 7.7 Which social class do you perceive you belong to? Poorest 66.1 19.7 25.2 8.1 7.0 4.9 1.7 2.4 Poor 49.2 47.2 35.4 36.5 11.1 25.2 4.3 19.3 Average 25.3 33.2 39.3 55.4 22.5 69.9 12.8 78.3

There is no significant change in families’ assessment of their economic performance in the past season although a minority in all groups reported an improvement. The bulk of the responses fall with a worsening of the situation and no change. Similarly gloomy have been the reporting of the family economic situation in the last year. The bulk of responses continues to fall within the no change and a worsening of the economic situation of the family. In terms of future expectations, however, only the worse off in society predominantly believe that things will not change or will deteriorate. They may have some objective reasons to believe this.

The bulk of the middle sector believes that their economic condition will not change. Within each cluster, the group that believes things will get better increases in size with higher income. The group of relatively rich people, which is cluster 6, is the most confident about the future with more than 50% believing the economic situation of the family will either improve a little or improve a lot. In view of the country’s economic performance the past two years, this confidence may look misplaced but as soon as one takes into account the pattern of income distribution of the past two years, more objective reasons for such confidence may arise. Finally, people’s perception of their own social situation is interesting at the top of the social strata. The majority of cluster 6, which accounts for less than 8% of the country’s population perceives themselves as just average, neither rich nor poor. Those at the bottom appear to have a better appraisal of where they are with 2/3 of them feeling poor or among the poorest.

Concluding remarks

We have offered in this technical annex a method for identifying characteristics of poor and non-poor families. This method is an alternative to lowering the poverty line, the latter an avenue of work that has dubious merits in facilitating the construction of a poverty profile in Georgia. The use of cluster analysis allows for a better identification of characteristics of the poor and non-poor at different locations in the income distribution curve. The method also allows avoiding a discussion of which poverty line is "more realistic" than other.

The results of Technical Annex 2 provide support to the findings of Technical Annex 1. For instance, the presence of high ranked officials as well as directors or managers exerts a positive influence on family welfare. This is clearly seen as these categories are over represented in the cluster that contains the richest strata of society. The clusters also show that only high government office or managerial position distinguishes very rich from the rest. While the presence of middle ranked managers is under represented in the poorest group, the middle and upper classes have a roughly equal representation. The clusters also confirm that skills are a necessary though hardly sufficient condition to go up the social ladder. In terms of skilled workers, their relative absence affects poor families but their presence fails to convey much ammunition to go up the social scale. All cluster except the lowest one have an equal number of them. The use of cluster is much more effective than variations in the poverty line to highlight the fact that well off families are likely to have a greater number of members with tertiary education than families down the social ladder. The use of cluster analysis, however, is not in itself enough to conclude that a tertiary degree is an effective tool to escape the ranks of the poor. In fact, for the specific case of Georgia, we suggest some caution in arriving to this conclusion. The results of Technical Annex 1 also suggest that the contribution of education to family welfare is dependent on the existing welfare level of the family and the presence of invisible characteristics like networks of contacts.